|Publication number||US7349838 B2|
|Application number||US 10/880,923|
|Publication date||Mar 25, 2008|
|Filing date||Jun 30, 2004|
|Priority date||Jul 31, 1998|
|Also published as||US20050004789|
|Publication number||10880923, 880923, US 7349838 B2, US 7349838B2, US-B2-7349838, US7349838 B2, US7349838B2|
|Inventors||Gary J. Summers|
|Original Assignee||Summers Gary J|
|Export Citation||BiBTeX, EndNote, RefMan|
|Patent Citations (30), Non-Patent Citations (50), Referenced by (42), Classifications (23), Legal Events (3)|
|External Links: USPTO, USPTO Assignment, Espacenet|
This application is a continuation-in-part of U.S. application Ser. No. 10/141,997, filed May 8, 2002, entitled “Management Training Simulation Method and System,” now pending, which is a continuation-in-part of U.S. application Ser. No. 09/364,280, filed Jul. 30, 1999, entitled “Management Training Simulation Method and System,” now allowed, and is also a continuation-in-part of U.S. application Ser. No. 09/364,489, filed Jul. 30, 1999, entitled “Management Training Simulation Method and System,” now U.S. Pat. No. 6,236,955, and which further claims priority under 35 U.S.C. § 119(e) from U.S. Provisional Application Ser. No. 60/094,900, filed Jul. 31, 1998, and 60/141,738, filed Jun. 30, 1999, the disclosures of which are hereby incorporated by reference in their entireties as if set forth herein.
The present invention relates generally to management training simulations (MTSs), which are computer programs or board games that help managers learn to manage and to understand business. More particularly, the present invention involves a computerized management training method and system that effectively teaches the development and use of knowledge and provides training in managing strategy, risk, innovation, and core competencies, as well as analyzing and correcting a manager's decision making processes and identifying a manager's unique judgmental biases and errors. It provides tailored, individualized training in managerial judgment and decision making.
MTSs are computer simulations that teach managers how to make better informed decisions. They present a manager with a lifelike situation simulated by a computer. The manager endeavors to improve the situation. To do this, he analyzes the situation and responds with a decision. The computer then calculates and displays the consequences of his decision. If the simulation closely approximates realistic situations, the manager learns how to confront those situations when they arise in the work environment.
MTSs are also called business simulations, business gaming, and business war games. Many business schools, corporate universities, consulting firms, training firms, and human resource departments use MTSs to teach a wide variety of subjects including marketing, finance, accounting, business strategy, supply chain management, and organization design.
There is a great need for this educational technology. People learn best from practical, hands on experience. Yet the primary source of such experience, one's business, is a difficult place in which to learn. Business experiments are not repeatable, decision consequences represent the outcomes of many influences, and the penalties for failure are potentially high. Business risks, costs, and complexity prevent a manager from engaging in the playful, mistake driven experimentation through which people learn best.
The predominant alternative to learning ‘on the job’ are books and classroom study. These methods are also limited. Applying intellectual knowledge to practice is extremely difficult. For example, no medical student is expected to move directly from Gray's Anatomy to surgery.
MTSs overcome the problems of learning ‘on the job’ and of classroom study. They are the ideal means for learning: experiments are repeatable while consequences are discernable and immediate. They condense years of experience into a few hours of study, thereby improving the learning that managers gain from their most limited resource—time. MTSs bridge the distance between intellectual understanding and practice (as cadavers do for medical students). They facilitate practical learning without risking “the patient”—one's career and company.
A manager will gain the following benefits by using MTSs to improve his management skill:
An important class of MTS within the general MTS architecture depicted in
To manage their firm and, specifically, to receive information and input decisions, students use an interface (205). This interface is typically an integration of the display and input devices shown in
The arrows in
Each application of an MTS is called a learning session. A learning session progresses through rounds where each round consists of the following sequenced steps:
Decisions need not be synchronized this way. Instead, the simulation can run continuously.
With an asynchronized simulation users need not submit their decisions at the same time. Users can make any decision and submit it to the simulation at any time. When the simulation receives a users' decisions, it incorporates them into its next manipulation of the simulated situation.
An asynchronized MTS presents students with an additional challenge. They must repeatedly query the simulation for data. They must do this to keep abreast of changes and actions taken by other users. Moreover, users must query the simulation and make decisions at a pace that “keeps up” with the simulation.
The following description focuses upon marketplace models and product design to facilitate the discussion of MTSs in general. MTSs require a marketplace model which represents both products and markets. MTSs typically require students to perform three tasks: (1) analyze the marketplace and competing firms, (2) design products and set prices, and (3) invest in business processes. The following describes generally how MTSs' represent products and markets and how they supply the structure required to facilitate the students' performance of their required tasks.
Products: Products in known MTSs generally include three types of product traits: business process traits, aggregate traits, and attributes. Business process traits represent the outcome of business processes, such as customer service level and delivery delays. Aggregate traits describe the whole product, such as product quality and product reliability. Attributes represent specific features comprising a product. Attributes can vary quantitatively (for example, amount of calories in one serving of a breakfast cereal) or qualitatively (for example, a product's color). The values that attributes can express are called characteristics. The set of characteristics that an attribute can express is referred to as the attribute's domain. The composite produced by the characteristics expressed by a product's attributes is called a product's design.
Product Classes: A product class is the set of products consisting of all the possible values for a product vector. Real world examples of product classes are sports cars and long distance phone service. A specific product is identified by its class and its traits. For example, suppose sports cars have three traits: customer service, delivery delay, and product quality. Suppose also that customer service and product quality are measured with a ten point scale. A specific product in the sports car product class is a sports car with a level five customer service, two week delivery delay, and a level seven quality.
To provide more realistic decision situations, some MTSs furnish several product classes, for example sports cars and luxury cars. Multiple product classes are defined by declaring their existence. For example, an MTS might declare three classes of products (classes A, B, C) by declaring three types of product vectors of the type described above. Each product class can have the same traits, but this is not necessary.
Markets: Demand for products is simulated in prior art MTSs using a demand function for a market manipulator (
It is notable that, usually, the market manipulator does not directly receive product characteristics as inputs (as independent variables). Instead, a product's characteristics are used to produce a single number that represents a market's evaluation of a product's design. I call this number a product's value. The conversion is produced by a product evaluator (
Management Decisions: Students are told what product classes, market segments, and markets exist and the product traits comprising the products of an MTS. With this knowledge, students control a firm and compete in the simulated marketplaces by producing products from one or more of the declared product classes.
Each student manages his firm by performing the following tasks:
The tasks of market analysis, competitor analysis, and investment in business processes are described below.
Market and Competitor Analysis: Students analyze the marketplace through three methods:
Students analyze competitors using two methods:
From a student's marketing and competitor analysis, he develops a business strategy. The business strategy states a focus on specific product classes, markets, and market segments. It states the desired values of product traits, prices, and production volumes. A student enacts his strategy with three decisions: set the attribute levels, set the prices of its products, and invest in business processes. These decisions are described below.
Setting Product Attributes and Price: Students set the characteristics expressed by their products' attributes. In setting characteristics, a student determines a product's design and is essentially designing a product in the simulation. The only restrictions on product design are the domains of the attributes. For example, quantitatively varying attributes might be bounded by minimum and maximum values. Likewise, qualitatively varying attributes might present students with a limited number of characteristics to choose from. Students also select their products' prices, subject to range limitations (for example, prices must be positive numbers).
Investing in Business Processes: Students improve their product's business process traits and aggregate traits by investing in their firm's business characteristics (for example, purchasing/scraping production capacity, retooling a factory, hiring new salespeople, or purchasing more advertising). The results are determined by equations that take the firm's characteristics and the student's investment decisions as the independent variables and yield the values of business process traits.
Equations giving a firm's characteristics can affect either business processes traits or firm characteristics, such as labor productivity. For examples of the use of equations in determining business process traits and firm characteristics, see: Steven Gold and Thomas Pray, “The Production Frontier: Modeling Production in Computerized Business Simulations,” Simulation and Games, vol. 20 (September 1989): pp. 300-318; Precha Thavikulwat, “Modeling the Human Component in Computer-Based Business Simulations,” Simulation and Gaming, vol. 22 (September 1991): pp. 350-359; Steven Gold, “Modeling Short-Run Cost and Production Functions in Computerized Business Simulations,” Simulation and Gaming, vol. 23 (December 1992): pp. 417-430; and Precha Thavikulwat, “Product Quality in Computerized Business Simulations,” Simulation and Gaming, vol. 23 (December 1992): pp. 431-441.
Modeling Innovation: The appendix provides a more detailed description of the prior art of MTS and also provides a general description of the prior art methods of modeling innovation, modeling technological advance, and the prior art product value functions.
Deficiencies of the Prior Art: The prior art MTSs suffer from six primary deficiencies:
Information about the value of products provides a lot of information about the value of all other products.
The present invention improves over the prior art by creating a new modeling relationship between a product's design and its value. The consequences of this change are great. Embodiments of the present invention can provide a superior model of innovation and technological advance, highlights the role of information and knowledge in management and in an industry's dynamics, and provides a means of explicitly representing a student's development and application of knowledge.
The foregoing brief description, as well as other features and advantages of the present invention, will be understood more completely from the following detailed description of illustrative embodiments, with reference being made to the accompanying drawings, in which:
By way of overview, the present invention improves MTSs by building upon new ideas from the fields of evolutionary economics, evolutionary biology, and genetic algorithms, and from studies in the management of innovation. Primarily, and with additional important consequences, the present invention improves MTSs' modeling of information, knowledge, and innovation. For the purpose of exposition, the sections titled “Introduction and Definitions,” “Illustrative Embodiment,” “Ramifications,” and “Embodiments that Make Two Great Improvements to MTSs” describe the specific class of MTSs called competitive industry MTSs. Those versed in the art will appreciate that the invention described herein also applies to MTSs that emphasize other business situations. Thus, the later section titled “The General Case” describes more general instantiations of the current invention. In addition, several sections further broaden the scope of this disclosure in addressing additional features, aspects, and issues that relate to simulations. These sections discuss embodiments of components on the invention (value functions and design objects), additional applications, a second illustrative embodiment, various hardware configurations, and flowcharts of some particularly useful embodiments.
An MTS, in accordance with the present invention, is a departure from prior art simulations in its use of attribute-characteristic representations of products, and the use of new product value functions. The use of product categories is also new. These concepts are introduced and defined below, and the consequences of incorporating these new features into MTSs are discussed thereafter in connection with an illustrative embodiment, ramifications, and two great improvements to MTSs.
Aspects of the invention are presented in eleven sections. The sections are as follows:
What follows is an outline and presentation of these sections.
Section 1: Introduction and Definitions
These sections are next presented, in order.
Products: Recall that some known MTSs describe products with attributes. For example, one may describe the design of an automobile with a list of attributes that includes physical qualities (such as color, size, and shape), features (such as antilock brakes and power windows), and abilities (top speed, city miles per gallon, highway miles per gallon). Using this representation scheme, automobiles are objects composed of the following attributes (style, engine type, drive train type, exterior color, window feature, brake feature, top speed, city mpg, highway mpg). Each attribute varies either quantitatively (for example, city mph and top speed) or qualitatively (for example style). Using a value function of a type described below, one can also have attributes that vary both qualitatively and quantitatively. An example of such an attribute is color. Color may vary qualitatively (for example, red, blue, and green) and in intensity (light blue to dark blue). One can represent intensity quantitatively with a number (for example, a ten point scale). I call an attribute that can vary both qualitatively and quantitatively a dual varying attribute. Including a dual variable attribute in the automobile example, automobiles are objects composed of the following attributes (style, engine type, drive train type, exterior color (intensity), window feature, brake feature, top speed, city mpg, highway mpg). A specific automobile is identified by the vector of characteristics (sports car, 4-cylinder engine, front wheel drive, blue exterior (intensity=5), . . . , electric windows, anti-lock brakes, 115 mph, 23 city mpg, 33 highway mpg). With this method, every product design is represented with a unique vector of characteristics. This method of representing a product's design is called an attribute-characteristic representation.
The attribute-characteristic representation is much more general than demonstrated by the preceding automobile example. The number of attributes in a product design can vary throughout a learning session and from product-to-product (like the way in which word length varies in a game of scrabble). Attributes can be real valued (such as top speed), integer valued (such as an integer scale of one to ten), or qualitative (such as a letters). In the case of qualitatively varying attributes, each attribute can express characteristics from different sets of characteristics (colors versus styles in the automobile example); characteristics from the same set, with duplications allowed (for example, letter combinations that produce words); or characteristics from the same set, without duplications (permutations). In addition, attributes can be diploid (such as dual varying attributes and the dominant-recessive genes made famous by Gregor Mendel's experiments with peas), triploid, or even more complex. Also, the attribute-characteristic method of accounting for a product's design can be recorded as a vector (as done above), with a matrix, as a single number, or through other suitable means.
The attribute-characteristic representation provides the means for representing all valid product designs. Recall that an attribute's domain specifies all of the possible characteristics that the attribute can express. The set of all valid product designs is produced by taking the cross-product of all the attribute domains (that is, taking every combination of characteristics).
The form of the attribute-characteristic representation is determined on a case-by-case basis with regard to the purpose of the MTS, the value function utilized (see below), and the available data structures.
Product Categories: A product set defined by product characteristics is called a product category. A notable quality of the attribute-characteristic representation of product designs is that one can readily define sets of products by characteristics. This quality is not important in prior art MTSs but, for reasons described below, is useful in the present invention. For qualitatively varying attributes, one defines a product category by the presence or absence of one or more characteristics. Three examples of product categories are (1) sports cars, (2) cars with four cylinder engines, and (3) sports cars with four cylinder engines. On the other hand, for quantitatively varying attributes, one defines a product category by specifying a range of values for the attribute. Three examples of such a product category are (1) cars with a top speed between 90 mph and 110 mph, (2) cars that have at least 20 city mpg, and (3) cars with a top speed between 90 mph and 110 mph and have at least 20 city mpg. In defining product categories, dual varying attributes are specified by the combining of the two methods illustrated above (for example, blue cars with color intensity between 3 and 7). A product category, therefore, can be defined based on the presence and/or the absence of attributes and can include any combination of qualitative or quantitative or other type of attributes.
The Product Value Function: The product value function assigns a value to each product. An important concept for understanding the kind of value functions used in the present invention is interactions. One can understand interactions by considering how attributes and characteristics contribute to a product's value. Interactions exist when an attribute's contribution to the value function depends upon the characteristics expressed by one or more other attributes. For example, how much value does a red exterior add to the value of a particular automobile? This question is difficult to answer. The value of a red exterior depends upon an automobile's style. It is highly valued on sports cars but not on limousines. In this example, the contribution-to-product value of a particular characteristic expressed by one attribute (here, exterior color) depends upon the characteristics expressed by other attributes (e.g., style). This effect is called an interaction and for some attribute-characteristics can be associated with “frustration.”
Frustration occurs when improving one attribute's contribution to product value decreases the contributions of other attributes. Strong frustration exists when the effect decreases the total product value and such frustration makes product design a difficult task. In the automobile example, changing the characteristic expressed by the style attribute from sports car to limousine increases the contribution of the style attribute to the product's value. Simultaneously, this change decreases the contribution of the red color expressed by the exterior color attribute. In total, the value of the automobile decreases.
As used herein, a product value function in which interactions produce strong frustration and therefore exhibit multiple otpima are referred to as “multipeaked value function” (see the glossary). The present invention uses multipeaked value functions in an MTS to more closely model, among other things, innovation. These functions can be found in optimization problems from a variety of fields, including combinatorial optimization, genetic algorithms, cellular automata, computer science, molecular biology, management science, and evolutionary biology. Specific examples of optimization problems that include a multipeaked value function include designing the layout of an integrated circuit, finding the shortest tour connecting a set of cities, scheduling production in a factory, and finding a protein that catalyzes a particular reaction.
A Visual Metaphor: One can understand the multipeaked value function, and its difference from the prior art distance value functions described in the Appendix using a visual metaphor. Consider all possible product designs to lie along a horizontal surface, with similar products lying close to each other and dissimilar products lying far from other products. Assuming no dual varying or similarly complex attributes, if a product has n attributes, one would need an n-dimensional space to properly accomplish this task. A visual metaphor is appropriate for considering a two attribute product.
To complete the visual metaphor, a mark is placed above each product at a height equal to the overall value of the product that it resides above. When this task is completed, one has created a distribution of product values over the surface of the products. This multidimensional arrangement of products together with the distribution of product values is called a product space. The shape of the distribution of product values in a product space is called the product space's topology.
The product space's topology shows how design quality varies over the product space. The product space can be visualized as a mountain range with the most valued products lying on the surface beneath mountain peaks and the least valued product lying on the surface beneath valleys. The topology of the mountain range significantly affects the nature of students' task in an MTS and the dynamics of an MTS. Does the product space topology resemble Mount Fuji: a single peak, with broad, smooth slopes? Does the product space topology resemble the Rocky Mountains: many peaks of varying heights, valleys of varying depths, and sudden changes in altitude?
In both the visual metaphor and the real multidimensional product space, the amount of frustration among attributes determines the topology of the product space. If there is no strong frustration (as in the prior art, where there is no frustration—see Appendix), the topology is Mount Fuji-like, that is, single peaked.
However, when strong frustration exists, the product space can have a multipeaked topology. In formal terms, a multipeaked value function has as least one local optimum, in addition to a global optimum.
In addition to its multiple optima, multipeaked value functions of the type used in the present invention possesses another property absent from smooth and some rugged value functions: an optimal product cannot be discovered by varying the attribute-characteristics independently. To illustrate this property, consider a product that has two attributes: a1 and a2, where each attribute expresses an integer between one and ten. Suppose that the product in the product space with the highest value has a1=3. In the prior art value functions, for any value of a2 three is the best choice for a1. Because of this, students using prior art MTSs can find the best product by treating each attribute independently (see the appendix). In the example above, once a student has discovered that three is the best value for attribute one, he need not consider attribute one again. He can focus exclusively on finding the best value for attribute two. However, when a multipeaked value function is used, the best value for an attribute depends upon the characteristics expressed by other attributes.
Correlation: Correlation is a function of the product value function and is closely related to product space topology. In important embodiments of the invention, students will use their knowledge of the value of one product design to predict the values of others and the quality (fit) of their prediction depends on correlation. The capability of information about the values of products to predict the values of other products is given by measures of correlation. For this reason, measures of correlation are useful in selecting the value function that produces the information properties desired for a particular MTS.
The predictive capacity of information increases and decreases with measures of correlation. These measures show that single peaked value functions are highly correlated over the product space. Information about the values of products provides significant information about the values of many other products. For multipeaked value functions, correlation decreases with strong frustration. Specifically, if changes in product design or product category definition include changes to attributes that exhibit strong frustration, correlation will decrease faster than in single peaked value functions. For an example of this effect, see: Bernard Manderick, Mark de Weger, and Piet Spiessens, “The Genetic Algorithm and the Structure of the Fitness Landscape,” in Proceedings of the Fourth International Conference on Genetic Algorithms, edited by Richard Belew and Lashon Booker (San Mateo, Calif.: Morgan Kauffman Publishers, 1991), pp. 143-150. As more or larger such changes occur, correlation will decrease even faster. Because of this quality, correlation on multipeaked landscapes is typically high for only small changes. Metaphorically, this means information is useful only in making localized predictions about the mountain range topology.
Students can use either of two methods to predict product values:
To calculate the predictive capacity of knowledge of the value of a given product for the purposes of a design change, one uses an autocorrelation formula with changes stemming from the existing product design. For a description of the autocorrelation function, see Manderick, et al., supra. The autocorrelation function is usually discussed with reference to qualitatively varying attributes. For quantitatively varying attributes, one generates the required sequences of designs by iterative applications of the following steps:
To measure the correlation between two sets of products, one performs the following procedure:
To aid an MTS designer in selecting a value function, there are several useful correlation functions discussed in related academic research, as well as their relationship to product space topology, see: Manderick et al., supra; see also Marc Lipsitch, “Adaptation on Rugged Landscapes Generated by Iterated Local Interactions of Neighboring Genes, “in Proceedings of the Fourth International Conference on Genetic Algorithms, edited by Richard Belew and Lashon Booker (San Mateo, Calif.: Morgan Kauffmann Publishers, 1991), pp. 128-135; Stuart Kauffman, The Origins of Order: Self-Organization and Selection in Evolution (New York: Oxford University Press, 1993), pp. 63-66. Bernard Manderick, “Correlation Analysis,” in Thomas Back, David Fogel, and Zbignies Michalewicz, Handbook of Evolutionary Computation (New York: Oxford University Press, 1997), section B2.7.3 (hereafter referred to as HEC).
Some Examples of Value Functions: The attribute-characteristic representation and the value function work together; however, in order to apply an attribute-characteristic representation, one must define a value function that can accept its form (that is its combination of qualitative, quantitative, and other, more complex attributes) or else the attribute-characteristic representation cannot be used in a model or simulation. “Management Applications and Other Classical Optimization Problems,” by Volker Nissen in section F1.2 of HEC provides a list of references of academic articles that investigate the optimization of these kind of value functions. From reviewing these articles, one can find many examples of suitable value functions, some of which may be found to be suitable to the task upon experimentation. Some examples of suitable multipeaked value functions for modeling innovation include:
Quantitatively Varying Attributes: One can create a multipeaked function by summing multiple single peaked functions, for example Gaussian distributions. The “after Fletcher and Powell” function described in Thomas Back's Evolutionary Algorithms in Theory and Practice (New York: Oxford University Press, 1996), offers another function for representing quantitatively varying attributes. In this case, each product attribute represents a coordinate axis. Using Back's formulas, one may incorporate any number of attributes. Such a function can be applied here to accommodate qualitatively varying attributes and dual varying attributes by coding one or more of the coordinate axes into a discrete representation. Genetic algorithms frequently convert a real numbered axis into a base two (bit string) representation. A description of this process is given in Alden Wright's, “Genetic Algorithms for Real Parameter Optimization,” in Foundations of Genetic Algorithms, edited by Gregory Rawlins (San Mateo, Calif.: Morgan Kauffman Publishers). Using an analogous process, one can convert a real numbered axis into a discrete representation of any base and with any number of attributes.
Qualitatively Varying Attributes:
In order to create dual varying attributes, in accordance with an aspect of the invention, one adds a reference point to the TSP. The intensity of any attribute is equal to the distance from the reference point. As students adjust the intensity of a dual varying attribute, the destination expressed by that attribute moves so that the new intensity equals the distance between the destination and the reference point. A product's value is then calculated with this new configuration of destinations. Note that with this method, any number of attributes can be converted from qualitatively varying attributes to dual attributes.
Selecting Good Products? What kind of products could the system just described represent? Building a simulation around a real product presents two problems. First, fitting a multipeaked value function to a real product's “value function” is problematic. Why should the real value function for automobiles be anything like a value function found in texts on genetic algorithms? Second, in many embodiments students seek the most valuable products. This means that students should not already know the topology of the multipeaked value function. Students might already know the value function of a real product; for example, most people already know that red is a bad color for limousines. When building an embodiment of this invention, one should create compelling products for which students do not know the topology of the multipeaked value function. The following are some useful examples:
A basic MTS embodying the present invention is described with reference to
Preferably this embodiment comprises two programs and a spreadsheet file:
The values of n, k, and b are selected to produce an appropriate product value function. As the number of product attributes, n, increases, the variation in product values decrease. For this reason, products in the illustrative embodiment preferably have fewer attributes and a greater number of characteristics. This allows for sufficient variation in product values while still presenting students with a sufficiently difficult design problem.
In the embodiment described herein, n=10, b=26, and k should have a value of 2≦k≦4. One can arrange interactions evenly over attributes. This produces a value function with high correlation for small changes in product design (for example, a change of a single characteristic) and low correlation for more substantial design changes. From these recommendations, one of skill may adjust the values of the n, k, and b parameters to suit their particular MTS needs.
The Market: In the marketplace model of this illustrated embodiment, products have only one trait: product value. A market manipulator 802 accepts product value as its independent variable and calculates demand each round. To calculate demand, the market manipulator preferably uses a set of equations as described in U.S. Provisional Application Ser. No. 60/094,900, filed Jul. 31, 1998, or the Gold and Pray system of demand equations. Steven Gold and Thomas Pray, “Simulating Market-and Firm-Level Demand Functions in Computerized Business Simulations,” Simulations and Games, vol. 15 (September 1984): pp. 346-363.
Technological Advance: At the start of a learning session, the illustrative embodiment restricts the domains of one or more attributes, thereby limiting product design to sufficiently low valued products (for example, only characteristics ‘A’ through ‘G’ are allowed in product designs). Students compete by designing products with the allowed characteristics. During the learning session, the restrictions are relaxed, either incrementally (a few characteristics each round) or altogether (all restrictions removed in a single round), as described below.
Royalties: Define products as similar if they differ by less than a predetermined number of characteristics. A product is new to the marketplace if (1) it is appearing in the marketplace for the first time and (2) no similar products have appeared in the marketplace. If a firm produces a product that is new to the marketplace it has rights to the product and all similar products. This means that, for a limited number of rounds, if competing firms produce the product or a similar product, they must pay a royalty to the inventing firm. The duration and size of the royalty are adjustable parameters, set at the start of the learning session. One versed in the art can readily set the royalty parameters as desired and/or to fit real-world industry practice.
Manufacturing: Each unit of production capacity is best suited for manufacturing a particular type of product. This is called its specificity. Production capacity's specificity is designated by a product design. For example, capacity of type ASDFGHJKLL is best suited for producing the product design ASDFGHJKLL. Producing any other type of product increases the variable cost of production. For example, let Z represent the number characteristics in which the design of a product to be manufactured differs from the specificity of the production capacity used in the production. Let Y be the base unit variable cost, and let d be a constant. The cost for producing each unit of product is:
Unit variable production cost=(d*Z)+Y
The value of d is set by the illustrative embodiment at the start of the learning session.
Production capacity is purchased/scraped in blocks of capacity (for example, one hundred units). The cost of a block is constant over all specificities and throughout the learning session. The same is true of the scraping value. Similarly, the base variable production cost is constant over all product designs and throughout the learning session. One versed in the art can readily set the manufacturing parameters as desired and/or in accordance with real-world examples.
In the illustrative embodiment, all firms are controlled by students. As shown in
In addition to production capacity, firms have three routines or methods. (Firms are programmed as objects in an object oriented programming language). These routines (1) update the firms' budget, (2) update the firms' product capacity, and (3) send the firms' products to the marketplace.
Each student has one interface. The interface has four fields for recording a student's decision. It also has routines for sending information to a student's firm (e.g. by posting the results from a form). (The interface is programmed as an object in an object oriented programming language).
Marketing Analysis: Each round, the marketplace results are recorded in the market database. Using charts, graphs, and/or any means that they deem appropriate, each student analyzes the market database.
Management: Based on his analysis of the market database, each student (1) analyzes the marketplace result, (2) designs products, (3) chooses products to manufacture, (4) buys and sells production capacity, and (5) determines production schedules. In steps 1 and 2, the student is trying to identify the best designs, the “peaks” of the value function. Referring to the mountain range metaphor, this act of iteratively designing and testing products to determine the best is called searching the value function.
The interface helps a student keep track of his decisions.
A student purchases/sells production capacity by increasing/decreasing the maximum production listed in the right column. The student can also purchase new production capacity by adding a new row to the list. This method must be used when purchasing production capacity with a specificity that differs from the firm's current production capacity. As a student purchases/sells production capacity, the interface automatically adjusts his firm's budget (displayed in field 1201).
A negative budget means that the student's production decisions require more capital than the student's firm has in the current round. If a student tries to ‘manufacture and ship’ with a negative budget as determined by a script, function call, or applet; in a conventional manner, the interface alerts the student to the problem. The student can then adjust his capacity and production schedule accordingly.
The Market Database Specifications
The market database is a spreadsheet file on a student's computer.
The Operation of the Illustrative Embodiment
This embodiment generally progresses through the five steps of prior art MTSs which are repeated each round. In addition, this embodiment also requires two additional steps that occur only once during a learning session. First, the illustrative embodiment initializes the learning session before the initial round. Second, the illustrative embodiment simulates a technological breakthrough during the learning session. Below, I describe the five steps repeated each round and then I describe the two additional steps.
Steps One, Two, and Three: The first three steps of the illustrative embodiment consist of each student (1) analyzing the marketplace information, (2) making decisions for his firm, and (3) sending these decisions to his firm through the interface. This includes the following tasks:
After completing these five steps, the round is complete, and the next round, if any, begins with step one. As the rounds iterate, students compete by searching the value function to find, and manufacture, the best products. For this reason, this embodiment is a competitive search of the product value function. Initializing the Learning Session: Before a learning session can begin, the computer must initialize the learning session. To accomplish this, it performs the following steps:
The MTS of the illustrated embodiment changes the simulation of innovation and technological advance. It also fundamentally affects the students' tasks of analyzing marketplace results and designing products. As these tasks are central to many MTSs, other tasks that an MTS demands of students are also affected, as well as the dynamics of the simulated industry. These ramifications are described below.
Designing Products: Students design products by selecting the characteristic expressed by each product attribute. When designing products, students face two problems. First, there are an enormous number of designs. In the described embodiment, for example, products can have ten attributes with each attribute expressing one of twenty-six characteristics. Thus, students can choose from 2610 unique products. A student can consider only a small number of these possibilities. Second, attributes interact and produce strong frustration. Because of this, students cannot optimize design by considering each attribute independently. Instead, each student must discover valued combinations of characteristics.
One strategy to efficiently design high value products is to hypothesize and evaluate product categories. By using product categories, a student greatly simplifies the design problem and learns decision making skills in the process. He can evaluate the potential of an entire category of products rather than evaluate every single product. Specifically, a student evaluates a category by observing the marketplace performance of a few products from that category. If the product category shows potential (its products fair well in the marketplace competition), the student concentrates his effort and investment in that category. If the category evaluates poorly, the student hypothesizes new categories that he believes will produce better results and implements those revised designs in subsequent rounds.
Projects: In order to develop valuable products more quickly and efficiently, the student hypothesizes several product categories and searches within each one. The exploration of each hypothesized product category is called a project. A student can manage a portfolio of projects, deciding when to initiate new projects, when to cancel projects, and how to distribute his firm's budget among projects. Selecting product categories to search is an important decision. The product categories that a student focuses upon define his business.
Perspective: Students need not randomly hypothesize product categories or randomly design products. Instead, they hypothesize product categories and design products after studying the marketplace results, for example, as may be provided in the marketplace database (see
A student faces a difficult problem in analyzing marketplace information. The marketplace produces an enormous amount of information whether attributes vary qualitatively, quantitatively, or both.
To cope with the voluminous information, a student must select the information that is most effective and relevant to his business. One strategy for accomplishing this is evaluating only a few product categories. These categories might include, for example, the student's projects, potential projects, and product categories defining his competitors' products (as defined by the student). I call this set of product categories a student's perspective.
A perspective has the effect of categorizing the marketplace data. In doing so, it filters the market information, selecting the information that a student feels is most important. It is the means through which a student ‘frames’ the complex problem of competing, surviving, and profiting. It can be interpreted as a student's definition of the market. Different perspectives filter the marketplace results differently. Students with different perspectives will identify and miss different opportunities; evaluate product categories differently; and value information differently. Results that are surprising to one student might easily be anticipated by a student with a different perspective.
Innovation: Because students design products, innovation is defined in terms of product design. Specifically, an innovation is a product that differs from the previous products offered to the marketplace by at least one characteristic. A student's innovations will come from his strategy of search. For example, every product category possesses a unique distribution of product values. When a student determines the product categories that he searches, he determines the distribution of product values corresponding to the innovations that he may produce. The student determines whether he searches a barren category or one pregnant with innovations. Moreover, a student changes the product categories that he searches as he gains knowledge. As a result, innovation is primarily a function of a student's development and application of knowledge—i.e., knowledge management. In general, a student determines and modifies his search strategy and thereby affects the opportunities before him. In stark contrast, innovation in prior art MTSs have relied upon exogenously determined sets of new product designs coupled with innovation probabilities and are essentially an investment decision where spending more increases the probability of designing better products.
Short- and Long-Run Strategies: Innovating via perspectives and product categories creates a dilemma for students. A student can direct his efforts and investment towards product categories that the marketplace results have identified as most promising (categories that have done well in previous rounds). By exploiting this ‘current’ knowledge, a student immediately increases his firm's profits and the competitive pressure on his competitors. This is a short-run strategy. Alternatively, a student can take a long-run strategy and invest in discovering new product categories that contain higher value products (that is, new core competencies as described below). By developing new knowledge, a student can gain a large competitive advantage in future rounds. This requires time and investment, and there is a risk that no such product categories will be found. Balancing investment between these two alternatives is the quintessential knowledge management dilemma.
A Continuum of Innovations, Incremental through Radical: Using either the autocorrelation function with a specified product as a starting point, or by measuring the correlation between product categories, one can define a measurable continuum of innovation types. Consider the product categories containing a significant number of the products offered to the marketplace in previous rounds. Incremental innovations are innovations from product categories that are highly correlated with at least one of these product categories. Radical innovations are innovations contained in product categories that are not correlated with these product categories. Innovation type is measured by these correlations.
Because the innovation measure can be defined as either changes in a product's design or as a comparison of product categories, this measure is relevant to students when they design products and study the marketplace results. If a new product is an incremental innovation, analysis of previous marketplace results provides a good prediction of the new product's value. Incremental innovation can rely primarily upon market analysis. In contrast, previous marketplace results are poor predictors of the value of radical innovations. Because of this, inventing a radical innovation requires testing new products in the marketplace. Compared to incremental innovations, they require greater time and investment. Their development also carries a greater risk of failure. With little guidance from previous marketplace results, students may not find any successful radical innovations.
Incremental and radical innovations have an obvious relationship to the problem of exploiting knowledge versus developing new knowledge with the short-run vs. long-run strategies. Investing heavily in incremental innovation is the exploitation of current knowledge. It is a short-run strategy. Investing in radical innovation requires developing new knowledge. It is a long-run strategy.
It is important to note that every product is a member of many product categories. A product presents a group of n characteristics to the market. The number of combinations of characteristics evaluated by the market is the number of sets that one can create from n objects. This number is 2n. Whether a student sees an innovation as incremental or radical, or to what degree in between, depends upon the student's perspective as well as on the new product. A student with a good perspective will be able to reduce the risks and costs of innovation.
Finally, it should be noted that by using correlation measures to choose a product value function, an MTS designer can change the relative number of incremental and radical innovations available for students when designing products. As value functions become less correlated, the MTS presents students with fewer incremental innovations and more radical innovations.
Technological Advance: A technological advance is simulated by restricting and then removing restrictions on the valid product designs. Depending upon the specific form of the attribute-characteristic representation, this may include restricting/expanding the domains of the attributes or the number of attributes. In the illustrative embodiment, restrictions on the domains of attributes in product designs are applied and later removed. After a technological advance, multitudes of new products and product categories become available to students. Students compete by exploiting these new opportunities. Metaphorically, after a technological advance students can search new areas of the mountain range. Formally, after a technological advance, students can search new volumes of the multidimensional product space. Students will have to develop new definitions of the market (new perspectives) and new definitions of their firms' businesses (new product categories). In extreme cases, students will have to ‘reinvent’ their firms. This requires developing new knowledge while shedding the knowledge made obsolete by the technological advance.
Measurability of Information: One of the important properties that can be developed in embodiments of the present invention is that information can be measured. There may be several measures, each useful for a different purpose. Two useful measures are the reliability of information and population statistics. In the case of qualitatively varying attributes the new method is combinatorial and is congruent with the mathematics of information theory.
To understand the measure of information reliability, consider the task of product design. In determining product designs and in hypothesizing product categories, students might use their knowledge of the value of products in a category to predict the values of products in other categories. To borrow a real world example, “What do strong sales of sports cars indicate about the values of utility vehicles?” Alternatively, students might use their knowledge of the value of a single product to predict the effects of design changes to that product. In either case, the reliability of information measures this predictive capacity. It is given by the appropriate correlation measure: the autocorrelation function or the correlation between two product categories.
The measure and usefulness of population statistics can be understood by considering the students' task of evaluating a product category. In the terms of statistics, the products that are evaluated in the marketplace competition are samples from a population (the products in a product category). The students' evaluation of product categories via marketplace results is similar to a statistician's evaluation of a population via sampling. Because of this similarity, population statistics apply to analyze how students choose products to send to the marketplace (how students sample the population). In addition, population statistics apply to provide objective measures of the values of products in a product category (for example, confidence intervals). These objective measures are compared to a student's subjective estimates to identify biases in the student's judgment.
Other Important Properties of Information: By modeling the product space in accordance with the invention, the illustrated embodiment possesses several significant properties of the information, including:
Product Categories: Hypothesized product categories determine the innovations a student might design (it is the product space where he looks). They determine—from a probabilistic viewpoint—the efficiency of a student's search for valued products. Stronger knowledge (product categories containing higher valued products) permits discovering (1) valued products with less investment or (2) higher valued products given the same investment. Through product categories, knowledge promotes efficient innovation.
Perspective: Perspectives select the information from the marketplace database that evaluates product categories. Comparatively, stronger knowledge (a perspective that includes product categories which differ greatly in their products' values) can separate high value product categories from low value product categories with less information or provide a better prediction of product categories' values given the same amount of information. Stronger knowledge (1) reduces risk because students have superior identification of high value product categories and (2) reduces the investment needed to find high value product categories. Through perspectives, knowledge reduces the risks and costs of doing business.
In both cases, stronger knowledge means being able to more efficiently focus one's resources to satisfy the marketplace.
Core Competencies: When searching a product category, a student learns the valuable combinations of characteristics for that category (the characteristics that the ‘#’ attributes shown in
One can record the development of a student's core competencies through statistical measures (measures of central tendency and variation) of the products that the student offers the market. For a set of products, one can measure a core competency with the vector (A1, A2, . . . , An, Var,
One can apply this measure of core competency to any set of products (for example, the products produced in a project, by a firm, or by all firms in a round of the learning session). By repeating this calculation over several rounds, one can track the evolution of core competencies.
Alternatively, one can define core competencies with correlation. Recall, from the discussion of projects, a student can hypothesize product categories and search for valuable products within the categories. To identify core competencies, pair every project with every other project. For each pair calculate the correlation between the two categories. Then apply cluster analysis to identify groups of categories that are highly correlated. Each group constitutes a volume of the technology space in which information is useful. Each group of highly correlated categories is a potential core competency. As a student searches a category and learns to design valuable products, he develops one of these core competencies.
Industry Life Cycles: Economists and technological historians have researched the birth, development, and demise of industries. They found that most industries develop through a three stage pattern called an industry life cycle (the automobile, commercial aircraft, and the minicomputer industries are just a few examples). James Utterback's book Mastering the Dynamics of Innovation (Boston: Harvard Business School Press, 1994) describes industry life cycles in detail (see chapter 4). As a brief review, industry life cycles progress as follows:
In accordance with the illustrated embodiment, a technological breakthrough can initiate an industry life cycle when either the domains of the attributes or the number of attributes is increased. Initially, the most valued products that firms can produce are restricted to have sufficiently low values. The students compete with these choices. Innovation will decrease as students find the most valued products in this limited set. Once this occurs, the domains of one or more attributes or the number of attributes is increased, thereby simulating a technological advance. This will initiate an industry life cycle.
Improvement 1: An MTS with Superior Modeling of Innovation
In a competitive industry MTS students guide a firm through a transition from a predecessor industry to a new industry. Such a transition simulates a technological breakthrough and concomitant displacement of an older technology (for example, the transistor destroying the market for vacuum tubes). By producing and using knowledge, students construct and adjust a portfolio of projects. Students must (1) define the market as the new industry develops; (2) build new core competencies and design new products; (3) protect against both short-run and long-run competitive threats; and (4) develop managerial rules appropriate for the industry's maturity (fluid, transition, and stable stages). Among other lessons gained by engaging in the simulation of a technological breakthrough, students learn the following:
Properties of Innovation: The process of innovation is unique among business functions. For descriptions of the characteristics of innovation and guidance on managing innovation see: Peter Drucker, Management: Tasks, Responsibilities, Practices (New York: Harper Collins, 1973) p. 782-803; Peter Drucker, Innovation and Entrepreneurship (New York Harper & Row, 1985) p. 143-176; Donald Frey, “The New Dynamism (Part 1),” Interfaces, vol. 24 no. 2 (March—April 1994): pp. 87-91; James Brian Quinn, “Managing Innovation: Controlled Chaos,” Harvard Business Review (May-June 1985): pp. 73-84; and Lowell Steele, Managing Technology: The Strategic View (New York: McGraw-Hill, 1989): pp. 263-288. The most startling characteristic of innovation is its unpredictability. The successful application and design of a radical innovation, for example, is rarely predictable at the start of its development. This unpredictability is the source of four other principle characteristics of radical innovations:
These rules contribute to success, but they are not sufficient. Implementing these rules requires judgment. Managers must determine which projects should be started; which, and when, projects should be cancelled; and determine when the firm should adjust its investment in a project. Managers must set aspiration levels; balance the risk of falling victim to a competitor's innovation with the risk of losing their investments in innovation; and negotiate the trade-off between flexibility and decision errors. With embodiments that have students make such decisions, students develop their judgment as they face these dilemmas in an MTS.
In learning these lessons, students will also confront and learn to manage, by interacting with the business situation through the user interface, the following issues:
Balancing the Risks of Lost Investment and Lost Opportunity: The risks of lost investment and lost opportunity are antithetical, as are the costs of their associated mistakes. How should a firm balance the current and future needs of the business? How does a firm maintain efficiency while also maintaining the flexibility that competition requires of the firm?
Managing in a Dynamic Industry: How much can a firm affect an industry's dynamics? How does one compare results to expectations when much of this analysis rests upon judgment? How does one evaluate a firm's wealth producing potential?
Portfolio Management: How many projects should a business pursue? How much diversity is advantageous, and how does diversity link to core competencies? What are good measures of innovative performance? How well do traditional financial calculations govern (for example, payback period, NPV, and ROI)?
Managing Change: How fast can a firm change its operating rules, core competencies, and product mix without endangering its survival? What kind of rules and measures result in change rather than stability? What rules effectively move resources from old opportunities to new ones? At what level of detail should one plan?
Improvement 2: Personalized Decision Analysis and Training
Prior art MTSs teach through an indirect method. A student tries various strategies, analyzes the results, and, hopefully, the MTS induces an improved understanding. This method of learning can be ineffective because a student learns only as well as he can invent strategies and induce lessons. In contrast, a direct method of teaching in accordance with another aspect of the invention analyzes a student's decisions and judgments in order to determine his unique, habitual judgment and decision strengths, errors, and biases. This cognitive analysis facilitates personalized training in critical thinking and business decision-making.
Potentially, MTSs are the ideal means of providing cognitive analysis and training. They present a student with well-defined problems and information that results in the receipt of well-defined answers. While necessary, these characteristics are insufficient. In order to provide personalized decision and judgment analysis, MTSs must meet two additional requirements. First, their design must facilitate measuring information and knowledge. Second, they must clearly relate the tasks demanded of students to cognitive functions that can be analyzed. Prior art MTS do not satisfy either of these two additional criteria. Embodiments of the present invention can satisfy both of them.
Means of measuring information and knowledge were described above. Embodiments of the present invention can clearly relate the tasks that they demand of students to cognitive functions. To understand the relationships, it is useful to recognize that in designing products students are actually competing to solve a complex optimization problem. Instead of using a scientist's powerful mathematical algorithms for this task, students use their own ‘cognitive’ algorithms. For example, embodiments of the invention can make students exercise three cognitive functions: covariation assessment, categorization, and judgment. The relationships between the students' tasks and these cognitive functions are described below.
Covariation Assessment: When students analyze the marketplace data, they are searching for correlations between combinations of product characteristics (product categories) and marketplace success. In cognitive psychology, this process is called covariation assessment. Experiments have tested peoples' covariation assessment in a variety of situations.
In one such experiment, subjects were shown several lists of paired variables and asked to estimate the correlation demonstrated in each list. Dennis Jennings, Teresa Amabile, and Lee Ross, “Informal Covariation Assessment: Data-Based versus Theory-Based Judgment,” in Judgment Under Uncertainty: Heuristics and Biases, edited by Daniel Kahneman, Paul Slovic, and Amos Tversky (New York: Cambridge University Press, 1982): pp. 211-230. This task is similar to students' analysis of marketplace information in the illustrated embodiment. In this embodiment, the paired variables are products and sales volume. The psychological experiment shows dramatic results. Subjects' estimates vary widely and, on average, greatly underestimate correlation. Correlations must be at least 0.8 before subjects, on average, estimate a correlation as high as 0.5. These results occur because subjects simplify their task by looking at only a few entries on the list. Correlation is a quality of the entire set, and only exceptional rows accurately convey this quality.
This study suggests that when market results do not make facts obvious, managers can be easily mislead by focusing their attention on a small set of information (for example, the striking success, the striking failure, firsthand experience, or benchmarking). In the illustrated embodiment, these errors can be recognized by the system, explained to the student in a report or other output, and corrected by the student to better avoid real-world errors. When market uncertainty exists, managers should rely more heavily upon decision rules and conduct a broad assessment of their firms' industry.
Another method that students might use to identify profitable product categories is to count the number of successful and unsuccessful products in a category. Psychological studies have also researched this method of correlation detection. Dennis Jennings, Teresa Amabile, and Lee Ross, “Informal Covariation Assessment: Data-Based versus Theory-Based Judgments,” in Judgment Under Uncertainty: Heuristics and Biases, edited by Daniel Kahneman, Paul Slovic, and Amos Tversky (New York: Cambridge University Press, 1982): pp. 211-230. With this method, the information available to students can be placed in 2×2 matrix, as shown in
Categorization: Categorization is a technique commonly used by people to simplify their environment. This is exactly what a student does when he hypothesizes product categories and a perspective (defining his business and the market). A student's categorization will have a dramatic affect on his performance. To see this, suppose that each student associates each project within his portfolio with an estimate of its potential for producing profits. This estimate can be represented as a probability and updated each round. Different categorizations will incorporate marketplace results differently. Because of this, students' expectations will evolve differently even though they view the same marketplace results. This will lead to different assessments of opportunities and risks and different actions. With an appropriately designed embodiment of the present invention, one can analyze how students form and change their categorization schemes by tracking the product categories and perspective used by the student in each round and how students' categorizations and other decision-making choices affect their management decisions.
Judgment: During the course of a learning session a student must make the following project management judgments: the value of the products in a product category; the costs and time required to find valuable products; and the reliability of information. The student must also judge his portfolio's risk, capital requirements, and potential for producing profitable returns. Finally, a student must also assess his level of confidence in his judgments. Each of these judgments can be input into the model through the user interface.
In each round of a learning session, one can solicit each of these judgments from a student. Furthermore, for each of these judgments, one can estimate the true value by calculating correlations between product categories and utilizing population statistics. From these values, the MTS administrator can identify which of the student's judgments are habitually erroneous. The administrator can also investigate how these errors affect a student's project and portfolio management. With a suitable definition of risk, one can perform an analogous analysis of a student's risk management.
Judgment analysis can address both a student's decision making and the impact of his decisions on his firm. It can address the following questions:
In addition to addressing these questions, an advanced judgment analysis identifies and corrects errors which are typical human thinking, such as biases from anchoring, overconfidence, honoring sunk costs, and scenario thinking. For a description of these biases see the appropriate chapters of: Robyn Dawes, Rational Choice in an Uncertain World (USA: Harcourt Brace Jovanovich, 1988); and Daniel Kahneman, Paul Slovic, and Amos Tversky, editors, Judgment Under Uncertainty: Heuristics and Biases (New York: Cambridge University Press, 1982).
The following two examples illustrate decision errors that an appropriately designed embodiment of the present invention could identify and correct:
Psychological studies of anchoring suggest that people overestimate the chance of success by as much as about 70%. This overconfidence can impact a firm by causing its managers to (1) bet on too few projects rather than building a diversified portfolio and (2) invest in projects long after development suggests that failure is nearly unavoidable.
Those versed in the art will recognize that one can embody the present invention and qualities described above in a variety of MTSs including the ‘general case’ MTS depicted by
This arrangement depicted in
This section illuminates the invention's broad scope. The following sections further illuminate the scope by presenting a multitude of configurations, examples, and applications. Sections 6 and 7 describe various configurations of value functions and design objects. Examples of many applications are given. Section 8 describes additional applications of the invention in modeling manufacturing, marketing, R&D, and technological advance. Section 9 presents another illustrative embodiment, which is a MTS for learning to manage a product development portfolio. Section 10 describes various hardware configurations, and Section 11 presents flowcharts of some useful applications.
Sections 6-10 present numerous examples of embodiments that are broader in scope than the “General” case described above. More generally, MTSs are simulated business situations in which users learn business, management, and decision-making skills by interacting with the system. MTSs are also used for diagnosis and evaluation, as described in section 9. The terms “MTS” and “Management Training Simulations” are to be understood as encompassing this broad definition, and not being restricted to the competitive industry or general case referred to above.
Furthermore, in sections 6-10 people using MTSs are described as users rather than students. The term “student” is meant to identify anyone using an MTS to learn, as defined in the glossary. However, the term “user” better denotes the concept, so the descriptions in sections 6-11 use the term “user.”
The presentation of multipeaked value functions discussed peaks, frustration, and interactions. To emphasize the role of interactions, an Interacted Value Function (IVF) is defined. To emphasize the role of “valleys” between local optima, an Interdependent-Attribute Value Function (IAVF) is defined. Both definitions appear in the glossary.
Further considering value functions, the multipeaked value function (MVF), as defined herein, has as its domain all products or objects in a given simulation. However, a value function need not apply to the entire set of object designs. A value function can cover only a portion of the designs. For example, a simulation might use several value functions, each covering a different set of design objects. A Value Function with Multiple Optima (VFMO) is a value function that has multiple optima and has as its domain a subset of the design objects in a simulation. The subset can be the entire set of design objects. (The domain of a VFMO need not be a proper subset of the set of design objects.) One of ordinary skill in the art should appreciate that the Selected Value Function (SVF), used in an MTS, can be a multipeaked value function, an interacted value function, a value function with multiple optima, or an interdependent-attribute value function and that the foregoing description should be understood as embracing all these types of value functions.
The nk-Landscape Function: Many value functions do not explicitly identify the interactions among attributes. The interactions are implicit in the function. The nk-landscape function explicitly identifies the interactions, and for this reason, it is useful for both illustrating and building MTSs embodying the invention. Because of its usefulness in illustration, it is described here and used in ensuing examples.
The illustrated embodiment uses an nk-landscape with ten attributes that can each express one of twenty-six characteristics. For the sake of demonstration, consider a simpler nk-landscape. Those skilled in the art can readily gain further description and illustration from The Origins of Order by Stuart Kauffman and section B2.7.2 of HEC. Suppose that each product design has four attributes that can each express a 1 or a 0. Suppose further that each attribute's contribution is affected by its neighbors.
The example is simple. The interactions are two-way: attribute 10 interacts with attribute 20 and attribute 20 interacts with attribute 10. All attributes interact with their neighbors, and each attribute uses the same data table.
A drawback of nk-landscapes is that the tables become large as the number of characteristics and interactions increase. Consider a design object where each attribute can express ten characteristics. Suppose that an attribute interacted with five other attributes. The table for this attribute would need one million entries (ten to the sixth power). Basically, as the number of interactions and/or characteristics increase, the size of the tables increases very fast. A complex nk-landscape function requires lots of computer memory.
There are two solutions to this problem. The first recognizes that much of the data in the tables is not needed. To see this, consider a design object with eight attributes that can each express twelve characteristics. There are nearly 430 million object designs. During a training session the simulation will only evaluate a tiny fraction of these designs, so most of the tables' entries will never be used. To take advantage of this fact, table entries should be created only when they are first needed. The entries are then stored for future use.
Unfortunately, this method can still require lots of memory, especially for long running simulations. In addition, it is cumbersome if one runs a simulation on a server and gives users access through client computers. Even worse, it is problematic if one transfers data between computers via a floppy disk or email. (See section 10 for a presentation of hardware configurations). Section B2.7.2 of HEC provides an alternative that overcomes all of these problems. It proposes a pseudo random function with a seed that designates the simulation. To evaluate objects' designs a computer need only the seed and the function (specified as code). There are no tables, because the nk-landscape function is completely coded as a formula.
Evaluating a Portion of a Design: A SVF need not evaluate the entire design. It can evaluate a portion of the attributes. The other attributes need not be evaluated, can be evaluated by another SVF, or can be used by the simulation for other uses.
Using Other Information in an Evaluation: A SVF can utilize information in addition to a design. Consider the example from
Multiple SVFs: An MTS can use several SVFs. For example, design objects can be products and an MTS can simulate several markets. Each market can have an SVF to assign values to designs. The same product design will have a different value in each market.
Changing the Value Function: A simulation can change a value function during a simulation. For example, it can use one value function during the first five rounds of a simulation, and then it can use a different value function for the remaining rounds. More common, a simulation will alter a value function during a simulation. Altering the value function has the effect of deforming the topology of the space that users search. These changes might simulate changes in macroeconomic parameters, government regulations, consumer tastes, or other “shocks” to the simulated business situation. In any case, a user's knowledge of the SVF topography becomes somewhat obsolete, and he or she must adjust decisions and designs to the new topography. Such changes can be automatically initiated after a prescribed number of rounds in the simulation; after one or more of a statistically determined percentage or certain specific design objects have achieved a prescribed value; in response to exceeding or failing to achieve a prescribed overall or specific change in the value of objects at a given point in the simulation; in response to the use of certain information by a user (e.g., after purchasing a report); or otherwise in response to a programmed criterion or multiple criteria across a number of rounds or users.
The methods used to change the value function depend upon the specific function used in a simulation. The following examples use value functions named earlier in this specification. From viewing these examples, it can be understood by one of ordinary skill how to alter or change whichever value function is being used in a given simulation.
Changes in Value Function (Example 1), The Traveling Salesman Problem: Section 1, Introduction and Definitions, introduced the traveling salesman problem (TSP) as a value function. Recall: “The traveling salesman problem is the problem of visiting each vertex (i.e., city) in a fully connected graph exactly once while minimizing a cost function defined with respect to the edges between adjacent vertices. In simple terms, the problem is to minimize the total distance traveled while visiting all the cities [in a set of cities] and returning to the point of origin.” Darell Whitley, “Permutations,” in section C1.4 of HEC, p. C1.4:1.
For the purpose of illustration,
During the simulation, the value function can be altered by changing the location of the vertices.
Changes in Value Function (Example 2), an nk-landscape function: During a simulation the simulation might alter an nk-landscape function by randomly changing some of the contributions (some of the entries in the right column of
The section “Fundamental Components” introduced objects that are described by an attribute-characteristics representation. In the examples given, the objects were products (automobiles). The attributes expressed physical qualities (color), technological performance (top speed and city mph), features (window and brake features), components (engine type), and abstract qualities (style). Each product had a value that expressed the product's desirability in a simulated marketplace. In addition, the presentation of the attribute-characteristic representation illustrated important capabilities of the method. Design objects can vary in length. Attributes can vary quantitatively or qualitatively. An object can have different types of attributes, some varying quantitatively, some varying qualitatively, plus other more complex types. In addition, useful, but more complex design objects can include segmented objects, system-set attributes, and objects with multiple values, as discussed in this Section.
Design Objects are More General than Products
While the “Fundamental Component” and “Illustrative Embodiment” sections usefully depict objects as a set of products, the concept of an object is much more broad. The design objects in a simulation can be any set of objects represented by an attribute-characteristic representation. For example, it could be advertisements which are sent to a market and whose success is then gauged.
Accordingly, the objects' values need not be market values (as in the illustrative embodiment). They can be any evaluation of a design object. In one particularly important case, the SVF assigns each object values representing technical performance. For example, consider an attribute-characteristic representation describing a laptop computer. The attributes are system components: battery type, screen type, processor type, operating system type, and others. The SVF determines the technical performance that arises from the interactions of the components. The technological performance includes: battery life, speed, heat production, and propensity to crash. In this example, the SVF does not determine market values. If desired, separate marketing equations determine market values from the technological performance (and possibly from other aspects of a simulation).
Using the attribute-characteristics representation and SVF to describe technology characteristics and performance (instead of product features and market values) is very powerful. For example, an attribute-characteristic representation can specify the design of a system and the SVF can determine manufacturing qualities of the design. A SVF might determine:
The “Fundamental Components” section defines object categories as a set of objects defined by object characteristics, particularly defined by the inclusion or exclusion of characteristics. The “Illustrative Embodiment” and “Ramifications” sections illustrated this concept by using “#” to represent any characteristic in the domain of an attribute. For example, (#####), (ABC##), are (ABCDE) are all object categories. Extending this notation enables one to fully express the notion of object category.
An object category is any subset of the set of design objects. One can broaden the notation presented earlier so that one can identify any subset by the inclusion or exclusion of characteristics. Let a single quotation mark signify the absence of a characteristic. For example, (ABCDE′) means all products with (ABCD) for the first four attributes and any characteristic except for “E” in the fifth attribute. Similarly, the category (####E′) means all objects that do not have an “E” expressed by the fifth attribute.
In addition, object categories can be defined using the union and/or intersection of categories. For example, letting “U” stand for union, the object category (ABC##) U (ABA##) means all products where the first attribute expresses “A;” the second attribute expresses “B;” the third attribute expresses either “A” or “C;” and the fourth and fifth attributes can express any characteristic. With this expended notation, one can define any subset of the set of objects.
Segmented Design Objects
Often, it is helpful to identify groups of attributes. Generally, a group of attributes can be associated with some aspect of the object, SVF, or another part of a simulation.
Segmented objects can have two types of interactions. Intrasegment interactions are interactions between attributes in the same segment. Intersegment interactions are interactions between attributes in different segments.
The concept of object categories applies to segments in addition to the entire object. Referring to
Segmented Design Objects and SVFs
When design objects are segmented, the simulation can use value functions in various ways. Consider the following examples:
When separate SVFs evaluate different segments the results need not be combined. The different SVFs can evaluate different aspects of the design object, and this is a means of assigning multiple values to an object. (Another method is described below). As an example,
When different SVFs evaluate different segments, the simulation need not evaluate each segment at the same time. In the medicinal drug example, users can focus their efforts on designing the first segment (drugs with strong healing effects). Later in the simulation they can design manufacturing segments specifically for the good healing effect designs (good segment 4210 designs).
There is one obstacle to evaluating segments at different times. Given the intersegment interactions, how can an SVF evaluate a segment if the user has not yet chosen the characteristics expressed in the other segments?
Second, suppose one wished to evaluate segment 4220, but that the simulation user had not yet selected characteristics for segment 4210. The simulation could evaluate segment 4220 for a variety of randomly generated segment 4210 designs. The simulation would then use a measure based on these results, such as the average.
SVFs Evaluating Overlapping Segments: Another arrangement includes overlapping segments.
The overlapping segments produce interactions. Suppose that the design object from
Examples of Using Segmented Design Objects
Segmented objects add new functionally to MTSs embodying the present invention. Below the following examples are described: market segments, teamwork, specialized knowledge, interfirm collaboration, pharmaceutical R&D, and product applications.
Market Segments: A segment of the design object can represent a product and another segment represent market segments.
For example, let (ABCDE) represent a product. Let (AAA), (AAB), and (BAA) represent market segments (which might represent teenagers, empty nesters, and senior citizens) or different product applications (for example applying glue to wood, plastic, or metal). Selling the product in all three markets produces three object designs: (ABCDEAAA), (ABCDEAAB) and (ABCDEBAA). Evaluating these designs with a SVF would produce three different values. In other words, the value of the product depends upon the segment that is buying the product.
Notice that the market segment attributes in
During a simulation, the users do not see the symbolic representation of the market segment attributes. In fact, they need not be aware of market segment attributes. Users simply select a segment in which to sell some of their products. The simulation then “attaches,” or concatenates, the symbolic representation for that segment. Alternatively, the computer can select the segment. An MTS might do this, for example, if it randomly generates customers entering a retail store.
Agent-Based Models: The illustrative embodiment presented above simulated a market. Design objects were evaluated by a multipeaked value function and the results were further processed by marketing equations. There are many systems of equations used to model markets in MTSs. For a review of equations for modeling markets see Steven Gold and Thomas Pray, “Modeling Demand in Computerized Business Simulations,” in Jim W. Gentry (ed.), Guide to Business Gaming and Experiential Learning, Association for Business Simulation and Experiential Learning (East Brunswick: Nichols/GP Publishing, 1990), pp. 117-138.
In addition to systems of equations MTSs can also simulate markets with agent-based methods. With this technology, which can build upon research from the field of artificial life, agents simulate individual customers or groups of customers. Products are sent to a simulated market where customers and suppliers conduct business. Agent-based models can make good use of segmented design objects. For example, one segment can represent a product design and the other can represent a customer's tastes. Each agent holds a copy of the attribute-characteristic representation of its tastes. The simulation can then calculate the value of a product for a particular customer. The product design segment is determined by the product in question and the customer taste segment is determined by the agent (customer) in question. To evaluate a design for a particular customer the product segment and the customer taste segment are concatenated and then evaluated by a SVF. In this example, the method is similar to using the segmented design to represent market segments, except that agents play the role of market segments. The agent-based modeling offers great opportunities for MTS. Many of these opportunities are enhanced by using segmented object designs and SVFs.
Teamwork: The simulation divides a design object's attributes into several segments (for example, attributes one through five, six through ten, and eleven through fifteen). Each user designs one segment. The intersegment interactions represent the impact of one user's decisions on other users' designs. To design high value products, the users must coordinate their efforts and work as a team. The users' dependence on each other increases with the number of intersegment interactions.
For example, consider the pharmaceutical example of
The simulation can further influence and teach teamwork through various scoring methods. For example, the MTS can evaluate each user's segment with a SVF and also evaluate the total design with another SVF. Each user's score is a weighted average of the values of his or her segment and of the total design. By changing the weights, the simulation can induce various levels of cooperation.
Specialized Knowledge: Consider the medicinal drug example of
Interfirm Collaboration: A simulation can have each segment represent a component of a product. For example, the product can be a personal computer and the components can be the motherboard, processor, hard drive, etc. Each user controls a firm, and at the start of the simulation each firm designs and manufactures a different component. In efforts to produce the final product, firms can purchase other components, form alliances, or diversify their product development. Considering the alliances, firms can design components together and/or jointly manufacture and sell products. When designing products together, firms must adjust designs for the interactions among components.
Pharmaceutical R&D: Pharmaceutical scientists design drugs that interact with proteins in the body. Greatly simplified, the process has two steps. First, a protein is identified through research of a disease or protein reactions such as cell metabolism. This protein is called a target. Second, scientists design drugs to interact with the target in a desired way. One can simulate targets with segmented objects.
Product Applications: An object can have one segment representing product designs and another segment represents customers' applications of a product (call these attributes application attributes). Intersegment interactions represent the effect of product changes on the application and the effect of using an existing product in a new application. This configuration is useful for simulating:
Sometimes, it is useful to let the simulation control one or more attributes or segments of a design object. When the simulation controls some attributes, these are called system-set attributes. Often, users do not see and are not aware of system-set attributes.
Some uses of system-set attributes include system shocks, multiple market segments, changes in consumer tastes, system components, and pharmaceutical R&D. Each of these is described below.
System Shocks: With system-set attributes MTSs can simulate shocks to the situation that they simulate. These shocks have the same effect as changing (or altering) the topography of value function (see the “Changing the Value Function,” section, above). In many cases, this is an easier means of achieving the effect.
Multiple Market Segments: Sometimes the users do not decide which market segment to send a product to. In this case, the simulation will calculate a product's value for the different segments by using system-set attributes. The method is similar to the previously described method for representing market segments (see above). However, in this case the user does not decide which market segment evaluates a product. The simulation makes this determination and then sets the system-set attributes. An MTS would do this, for example, when simulating patients' reactions to drugs. The system-set attributes represent different population segments of patients.
Changes in Consumer Tastes: Suppose that the user designs cell phone covers. The system-set attributes can represent factors influencing market tastes. By intermittently changing the characteristics expressed by these attributes, the simulation simulates changes in consumer tastes. The simulation changes the characteristic of a weakly interactive system-set attribute (
System Components: In another example, the users see the system-set attributes and know what they represent. However, the users cannot change the system-set attributes. Suppose that users compete at designing a component in a complex system. The system-set attributes represent the other components of the system. For example, consider a personal computer. The system-set attributes represent the motherboard, processor, devices, and other components. The users must design an operating system to work well with them. Periodically, one of the other components changes. Some changes are small, for example an improved protocol for communicating with CD-ROM drives. The simulation simulates this by changing the characteristic expressed by a weakly interactive system-set attribute. Alternatively, personal computers can change processors. The simulation simulates this by changing the characteristic expressed by a strongly interactive system-set attribute. In both cases, the users must adopt their design to the new system configuration. Note: users need not see symbolic representation of designs. Users can see written descriptions or visual depictions of the other system components. In the current example, the simulation might display a picture of a motherboard with installed components.
As is common in the personal computer industry, users can receive beta versions of the new components before the switch. The user can create designs for the new component. However, the beta version differs slightly from the component that is later offered to the market. An MTS can simulate this situation with two system-set attributes. One attribute has many interactions with the design attributes. The characteristics of this attribute represent the different types of components (for example, processor types 1, 2, and 3). The other system-set attribute has few interactions with the design attributes. This attribute has several characteristics. These characteristics indicate the versions of the component, for example beta version, version 1, version 2, and so on.
Research and Development: Recall the description of pharmaceutical R&D. Once scientists design a drug to interact with a protein, the drug is tested in a laboratory, in animals, and then finally in human beings. The simulation can use system-set attributes to represent the different tests on a drug. The system-set attributes enable the simulation to calculate different results for the same drug design, depending on the type of test.
Design Objects with Multiple Values
Previously, I showed how to assign multiple values to design objects using multiple SVFs. One can achieve the same effect using a single SVF and either a segmented design object or system-set attributes. The following example illustrates. Suppose that the user must design a “fountain of youth drug.” The drug is a string of five molecules, where each molecule is chosen from a set of nine possibilities. These possibilities are designated by the letters “A” through “I.” This makes the attribute-characteristic representation of the drug a string of five design attributes that can each express a letter “A” through “I.” There are 59,049 different designs (nine to the fifth power).
Suppose that “fountain of youth” drugs have five effects on patients:
The simulation uses the two effects attributes, E1 and E2, to represent each effect. Since there are five effects, the characteristic combinations are labeled (AA), (AB), (BB), (AC), (DD), each designating a possibility (AB means E1=A and E2=B). The simulation uses M1 to represent the markets. Specifically:
To set a value for an effect of a drug design, the simulation combines the system-set attributes with the user's design. Suppose, for example, that the user designs a drug (EFGHJ). To calculate the drug's Weight effect on a man the simulation evaluates the design (EFGHJABA) with the SVF. To calculate the drug's Heart Damage effect on a woman the simulation evaluates the design (EFGHJACB) with the SVF. In this way, the simulation uses system-set attributes to evaluate different effects of designs for different market segments.
With nine characteristics available for each attribute, the simulation could have represented all five effects with only one attribute that varies from A to E. Using two attributes allows the simulation to control the relationship between effects. The attribute E1 can have high interactions with design attributes, and E2 can have low interactions with design attributes.
With the interactions depicted in
The preceding sections describe how the use of the attribute-characteristics representation and SVFs creates MTSs with novel and outstanding properties. It can also produce equally novel and outstanding changes in the modeling of manufacturing, marketing, R&D, and technical advance. Section 8 describes these innovations.
No prior art assigns an attribute-characteristic representation to production capacity. Instead, the prior art primarily considers modeling production functions. For examples see Steven Gold's and Thomas's Pray's paper “The Production Frontier: modeling production in computerized business simulations,” in Simulation and Games, vol. 20, no. 3, September 1989, pp. 300-318. Also see Steven Gold's paper “Modeling Short-Run Costs and Production Functions in Computerized Business Simulations,” in Simulation and Games, vol. 23, no. 4, December 1992, pp. 417-430.
Assigning an attribute-characteristic representation to production capacity enables a simulation to teach coordinating manufacturing with firm strategy. It makes manufacturing an important consideration in marketing, R&D, and other business decisions, as described below. The method readily adapts to a variety of situations that an MTS might simulate. The drawings and ensuring descriptions make this apparent.
Modeling Manufacturing, description of the method:
One can assign an object category to a production capacity. Object categories are defined with an attribute-characteristic representation, and they are described in the Section 1, “Introduction and Definitions,” in the subsections titled “Products” and “Product Categories.” They are further described in Section 7, “Design Objects.” The object category assigned to a capacity is called the capacity's specificity. The previously described illustrative embodiment taught assigning to production capacity an object design defined by an attribute-characteristic representation. An object design is actually an object category. It is a category that contains one object: the design itself.
Considering the other parts of
Modeling Manufacturing, operation of the method:
A simulation can calculate the capacity's properties (step 4840) through a variety of methods. For example, production capacity with specificity defined as a large object category is more expensive. Capacity with specificity (AB###) costs more than capacity with specificity (ABC##). Production capacity with specificity (#####) is the most expensive. Capacity defined by an object design, for example (ABCDE), is the least expensive. Efficiency can be similarly defined. Efficiency decreases with the size of the object category. The capacity (#####) is the least efficient, and capacities defined by an object design, such as (ABCDE), are the most efficient.
Alternatively, the simulation can use a SVF to assign cost and efficiency. Capacities are then defined as object designs (object categories containing a single design), and their designs are inputs for a SVF. The SVF can calculate multiple qualities as described in the section titled “Design Objects with Multiple Values.” With this arrangement, the user must design specificities that have lower costs and higher efficiencies. The user must engage process innovation. The complexity of the SVF and the user's knowledge of the SVF determine the difficulty of designing production capacity for a particular product.
While the cost of production capacity is calculated at the time of purchase, other manufacturing qualities are determined when products are manufactured.
A simulation can calculate the qualities of production (step 5040) through a variety of methods. Consider the following three examples. For each example, suppose that the user orders 300 units of product (ABCDE) and 100 units of product (BBCDE). Suppose that production has two qualities: per unit cost and defect rate.
In the first example the user intends to manufacture the products using production capacity with specificity (ABC##). The simulation determines if the products are members of the object category (ABC##). If the product is a member, the production cost is $1 per unit and the defect rate is zero. If the product is not a member, the production cost is $1.50 per unit and the defect rate is 10%. The product (ABCDE) is a member, so producing 300 units costs $300. There are no defects. The product (BBCDE) is not a member of the object category (ABC##). Producing 100 units costs $150 and produces 10 defects. Only 90 products are fit for sale.
In the second example the production capacity's specificity is (ABCDD). The simulation calculates per unit cost and the defect rate by comparing a product's design to the capacity's specificity. It counts the number of characteristics that do not match. The unit cost equals $1+$0.5*(# of nonmatches). The defect rate is 0.1*(# of nonmatches).
In the third example, the production capacity's specificity is (DDCCE). The simulation calculates per unit production cost and defect rate using a SVF. The simulation starts by concatenating a product's design with the capacity's specificity. This produces the sequences (ABCDEDDCCE) and (BBCDEDDCCE). These sequences become inputs to SVFs. One SVF evaluates the concatenated designs to give the unit cost, and the other SVF evaluates the concatenated designs to give the defect rate. Alternatively, a single SVF can calculate both qualities using the techniques described in the section “Design Objects with Multiple Values.” This third example is particularly useful in MTSs in which the user designs new products. He or she must also design the manufacturing capacities for the new product.
Modeling Manufacturing: some variations of the method: Those skilled in the art will recognize that the method supports a great variety of configurations. Here are some examples:
Specificity: In the examples above the attribute-characteristic representation defining production capacity has the same number of attributes and the same set of characteristics as the attribute-characteristic representation defining products. This need not be the case. Furthermore, being an object category, the specificity can vary in all of the ways that the attribute-characteristic representation can vary and all the ways that object categories can be defined.
Affected Qualities: The affected qualities can be virtually any quality related to manufacturing. They can be manufacturing qualities, such as per unit costs, defect rates, energy requirements, raw material requirements, yields, throughput rate, and time required for manufacturing. They can be qualities of production capacities, such as cost, maximum production rate, and products that the production capacity can or cannot manufacture. The affected qualities can also be qualities of manufactured products, such as quality, reliability, or weight. The affected qualities can be product attributes that change their expressed characteristic during a manufacturing process. The affected qualities can even be attributes in the definition of the product capacities, so that the act of manufacturing changes the specificity of the production capacity.
The affected qualities can also be qualities of objects that are not directly involved in manufacturing or objects that are not involved in manufacturing. For example, the affected qualities can be qualities of a firm, such as worker safety or worker morale. The affected qualities can be environmental qualities, such as pollution resulting from manufacturing. The affected qualities can also be qualities of customers, such as the propensity to purchase. (Think of customers smelling the aroma emanating from a barbeque restaurant.)
Affecters: The examples above illustrate the use of a product's design and a capacity's specificity in calculations of manufacturing qualities. Obviously, an infinite variety of calculations can use these inputs. The calculation can also include additional inputs, such as the age of the production capacity, the variety of products being produced, or the products previously produced by the capacity.
Operation: One may implement the method in a great variety of ways. For example, instead of letting the user define a capacity's specificity the user can select from a list of available production capacities. This can represent buying manufacturing equipment from a vendor's product line or buying manufacturing capacity from the “second-hand” market.
Other Production Processes: The production process need not be manufacturing a product. For example, it could be manufacturing materials for use inside the firm, such as building prototypes or laboratory equipment used in R&D.
Components of Manufacturing Processes: Other known simulations represent the manufacturing process itself, with assembly lines, work stations, and work in progress inventory. These models are often made with discrete simulation or system dynamics technologies. In this case, one can assign one or more factory components (e.g., assembly lines, machine, and/or work stations) a specificity defined by an object category. The affected qualities can be the qualities of the components (speed of operation); of the components' operations (amount of scrap, production of heat); or of work in progress inventory (number of defects). As the qualities propagate through the simulated manufacturing system, these can influence “downstream” processes and the overall manufacturing process. They can affect queue lengths, throughput, machine breakdowns, and the flexibility and efficiency of the system.
Modeling Manufacturing, ramifications: Generally, the new method improves business simulation by making manufacturing management essential to business strategy. Manufacturing becomes an important consideration in marketing. Modifying a product affects manufacturing, possibly increasing costs or decreasing product quality. Moreover, changing production levels, extending a product line, or discontinuing a product can adversely affect production. Marketing decisions might require additional investment in production capacity. Because capacity is an unrecoverable “sunk” cost, buying capacity is a risky decision.
The user must also consider manufacturing when making R&D decisions. In a simulation, technical advance can make products obsolete. Therefore, technical advance can render a firm's production capacity worthless. Also, a user might have to discover the best specificity for decreasing manufacturing cost or improving some other affected quality. He must engage in process R&D.
By affecting other qualities of the firm (employee morale) and the firm's environment (pollution), a user must consider manufacturing in making other decisions for his firm. In all of these ways—affecting marketing, R&D, and other decisions—the method firmly integrates manufacturing into the user's strategic decision-making.
While the previous description of modeling manufacturing offers specific examples, these should not be construed as limitations on the scope of the invention, but rather as an exemplification of some embodiments. Accordingly, the scope of the invention should be determined not by the embodiments illustrated, but by the appended claims and their legal equivalents.
In General Terms: The method is can be used to model other business processes, in addition to manufacturing. Referring to
The examples above illustrate how an adjustor might affect affected qualities when design objects are being processed. This is useful for illustrating models of manufacturing. Those versed in the art will recognize that similar calculations can be made when object categories are being processed by resources. Processing object categories is useful for modeling business processes such as R&D. Here are four examples of comparing object categories to resources' specificities and then adjusting affected qualities.
First, a previous example illustrated a resource defined by an object category and objects to be processed defined by object designs. One can repeat the example where resources are defined by object designs and the objects to be processed are defined by object categories. As a second example, suppose a resource has a specificity (ABC##) and it processes object categories (BBC##) and (ABC#E). The category (ABC#E) is contained in category (ABC##). This is because all of the design objects in (ABC#E) are in (ABC##). The object category (BBC#E) is not contained in (ABC##). Therefore, the adjustor will decrease the cost of processing (ABC#E) and increase the cost of processing (BBC#E). As a third example, the adjustor can compare object categories with a resource's specificity and count the number of attributes expressing the same symbol. Consider the resource and object categories in the previous example. The specificity (ABC##) and (BBC#E) match three of five symbols. The specificity (ABC##) and resource (ABC#E) match four of five symbols. Processing the category (ABC#E) will cost less than processing category (BBC#E). As a fourth example, the adjustor can count the number of nonmatches occurring in attributes not expressing “#.” For example, (ABC##) and (BBC#E) have one nonmatch, while (ABC##) and (ABC#E) have zero nonmatches. For further illustration, (ABC##) and (###DC) match zero nonmatches. The cost of processing can increase with the number of nonmatches. Clearly, those versed in the art can conceived of various methods for an adjustor to compare object categories to resources specificities and then adjust affected qualities.
Recall that markets can be modeled with systems of equations or with agent-based methods. Agent-based methods can take exceptional advantage of the attribute-characteristic representation of objects, for example by exploiting segmented objects as previously described. Whichever model is used, the results are that object designs are evaluated with a SVF and the results are used in a simulation of a market. The term “market model” covers both system of equations and agent-based models of markets. Using this term, design objects are evaluated by a SVF and the results are used in a market model.
In prior art, MTSs users “conduct” market research by buying reports. These reports typically list the products sold and the sales volumes, prices, and market shares of each product in each market segment. These reports are generated by the MTS, so users do not actually conduct market research. Simulations using agent-based market models enhance this method by enabling users to survey and/or “interview” the agents representing customers. One embodiment of the present invention improves upon the prior art by integrating the attribute-characteristic representation of objects into the prior art methods of modeling market research.
Generally, whatever the market model, the market data is recorded by the MTS, perhaps being held in a database. Users can query this data. One unique capability that can be included in an embodiment of the invention is defining database queries with object categories and/or receiving search results that are categorized by object categories.
Many types of queries are possible. For example, queries can be defined on any of the following criteria: a product (object) category; a range of rounds (or time) in the simulation; one or more designated market segments; and one or more designated firms; and sales performance defined on volume, revenue, or market share. The results could include a wide variety of information including the following: market share per round; sales per round; total sales in all rounds; and a market segment's percentage of all sales in the simulation. One skilled in the art can think of other query parameters and results.
With agent-based market models embodiments of the current invention provide new capabilities. Users can query agents with queries defined on object categories and/or receive query results that are categorized by object categories. For example, a user can survey the market to learn how the market values a particular category of products. The user submits the product category to the MTS. The MTS then randomly generates several products from this category and randomly selects several agents (i.e., customers). Each agent evaluates each of the randomly generated products. All of the evaluations are averaged and the result is returned to the user. Notice the obvious fact from population statistics. As more products are evaluated by more agents, the estimate becomes more accurate. An MTS might have the user set the investment in the market survey. A greater investment increases the number of products and agents sampled.
Whatever the market model, the key to the new method is that either queries are defined on object categories or the query results are defined on or presented as object categories. Recall that an object design is also an object category. Note also that queries to a database of market information extract information that was determined by evaluating an object design with a SVF. Usually, the evaluations are subsequently processed with marketing equations. Queries to agents in an agent-based model have to generate the information to be returned, and this generation requires evaluating at least one object design with a SVF. In one case the evaluation of the object design occurs before the query and in the other case the evaluation occurs after the query. In either case, the information returned is derived from the evaluation of at least one object design by a SVF, even when the evaluation is subsequently processed by marketing or other equations.
Recall the previous discussion on object categories in Section 7. When modeling market research with this method, object categories can be defined using any of the methods described in Section 1, “Introduction and Definitions,” and in Section 7, “Design Objects.”
The term “R&D” refers to the process of developing new technologies and products. It covers everything from basic science, to product development, to modifying an existing product. An example of basic science is pharmaceutical research to create new drugs. An example of product development is designing the next generation of a microprocessor. An example of altering a product is adjusting the tread pattern of a tire to improve traction when driving in snow. The term “R&D” even includes work incidental to technology and product development. For example, it includes process R&D, which is research and development of manufacturing methods for new technologies and products. As another example, it also includes managing the R&D process, such as reviewing projects and managing a portfolio of product development projects.
The attribute-characteristic representation and the SVF enable sophisticated simulations of R&D. It is particularly useful when using the attribute-characteristic representation to represent technology and the SVF to assign technological performance. One represents R&D in the simulation by permitting users to search the SVF for good designs. In some simulations the user is searching for designs with high values, and in other simulations the user is searching for designs with low values. Still, in other simulations the user can search for designs with values close to some desired value. Generally, the term “good” refers to the user's goal when creating designs.
A user might search with two general methods, either in combination or exclusively:
In both cases, the manner in which a user submits a design or category (that is, sequentially or in groups) can pertain to the actual submission or to various measures of time (e.g., simulation rounds). Likewise, the information returned to the user can be returned immediately, at the end of a round, several rounds in after the user makes the submission, or according to another scheme.
The information returned to the user can vary with the purpose/design of the simulation. For example, if the user submits designs the simulation might return the actual design value, or it might simply indicate if the design's value exceeds a threshold value. Alternatively, the simulation might indicate whether a design is more or less valued than a previously submitted design or other benchmark. In any particular simulation, the response to such submissions is prescribed by a rule or algorithm.
Since a category identifies a set of objects, submitting an object category permits returning additional types of information. The simulation might return estimates of the average value of designs in the category, the variance of values for designs in the category, the highest/lowest value of designs in the category, the percentage of designs in the category with values exceeding a threshold or benchmark, or some other prescribed statistical determination.
When designs have values for several effects/markets (see the section “Design Object with Multiple Values,” above) the simulation can return information about one or more effects/markets. In this case, and all others mentioned, the simulation can determine which information to return and/or the user can request the type of information returned.
R&D Costs: Under any scenario, information requests might be costly, and many cost schemes are possible. When submitting individual designs, the user might be charged for each submitted design and for the type of information requested. The user might also be charged different amounts for different information. In the “Fountain of Youth” drug example, information about the “Weight” effect might be costly while information about the “Face” effect might be cheap. The difference represents the costs of different R&D tests. Similarly, a test that determines if a design's value exceeds a threshold might cost less than a test that returns the design's value. The cost might depend on other factors, such as the number of characteristics by which a submission differs from previous submissions.
The category case presents additional cost schemes. A user might be charged for each category submitted or a cost that depends upon the number of designs contained in a category. The user might be allowed to select the accuracy of information received about a category. If the simulation returns an estimate describing the values of designs in the category (e.g., average value) then the larger the number of samples, the greater the cost. The user might select the designs used in the sample. Alternatively, the simulation chooses them on a prescribed basis, say, randomly.
In another variation, the cost is a function of the characteristics of the submitted design, category, or of the characteristics of the samples used to evaluate a given category. For example, the function represents the cost of purchasing, producing, using, or assembling characteristics. So, as an illustration, if the design represents a personal computer, some components are more expensive than others to purchase, install, and/or test within the system. Sometimes testing two components is more expensive when the components are used together in the same design, as opposed to testing them separately.
Reading across the first row of
The simulation receives the request, deducts the costs from the user's firm, and processes the request with a SVF. The simulation returns a form with a table like the one depicted in
An MTS can simulate this quality of R&D by having a user purchase R&D resources. The resources can be suited only for testing a specific design or category of designs. Testing other design categories and designs is either infeasible or incurs penalties. The penalties can include, for example, increased cost, decreased accuracy of results, or needing greater time to perform the tests. To avoid penalties the user must purchase new R&D resources that are suited for designs and categories that will be tested. Additionally, R&D resources might have a capacity that limits the number of R&D tests in a given duration (e.g., a simulation round). Users must purchase additional resources to increase their R&D capacity.
The current method of modeling R&D can also model R&D resources. It uses the same method described above for representing manufacturing capacity. Recall that the method of modeling manufacturing included production capacities, affected qualities, and an adjustor. The adjustor changes affected qualities based upon the specificities of production capacities. When calculating the adjustments the adjuster can also consider the design of products being produced by the production capacity. When modeling R&D resources the R&D resources play the role of production capacities, and they are assigned specificities defined with design categories. They can also have capacity that limit the number of “searches” that they can perform in a give time period. The affected qualities can be any quality related to R&D, such as the cost of R&D resource, the cost of searching the SVF, and the accuracy of the tests. The object categories and object designs that the user submits to search the SVF play the role that products do in the method of modeling manufacturing. As with the method of modeling manufacturing, the adjustor does the work. It considers the specificity of the R&D resource and affects the affected qualities. In some cases it will also consider the object designs and object categories to be processed. To illustrate this method of modeling R&D resources,
A noteworthy case occurs when the user submits an object category to be evaluated and the simulation evaluates the category by evaluating numerous object designs from the category (described above). Each design that is evaluated can be compared to an R&D resource's specificity. Affected qualities are adjusted based upon these comparisons.
Patents: Optionally, a given simulation can permit a user to apply for a patent on a design so as to prevent other users' firms from benefiting from that particular design. Such a simulation emulates patent law by barring patent protection or allowing it depending upon, for example, whether a design was described in a prior report that was published or issued to another firm in the simulation, or sold commercially prior to requesting patent protection, or on the basis of other real-world rules. The scope of any simulated patent is defined by the particular design submitted for patent protection by that user. The scope can cover other designs that differ in one or more characteristics, if the simulation so permits. The simulation includes, in this embodiment, rules to decide when a given design is within the scope of a firm's patent. For example, a patent might cover a particular design plus all designs that differ only by the characteristic expressed by a particular attribute.
Segmented Design Objects: The method of modeling R&D can use all of the techniques described in Section 6, “Value Functions,” and Section 7, “Design Objects.” The technique of segmented design objects is particularly useful, especially when this technique is used to assign multiple values to objects. The simulation can let a user research each segment by itself, rather than tackle the entire design at once. For example, a user can research good designs for the first segment, ignoring the second segment. The user can then research good designs for the second segment, ignoring the first segment. Finally, the user can combine the segments and then make design adjustments in response to the affects of intersegment interactions. Other arrangements are possible, as suggested in the preceding section on segmented designs.
Alternatively, firms can develop specialization (e.g., core competencies) in certain segments, and a firm can seek to patent its new-found knowledge and prevent other firms from utilizing the patented segment(s). A rule base can emulate existing laws to bar patent protection or allow it depending on, for example, whether the segment was described in a prior report that was published or issued to another firm in the simulation, or sold commercially prior to requesting patent protection.
Reverse Engineering: The method of modeling R&D can also model reverse engineering. To illustrate, suppose that several users compete in a simulation of the pharmaceutical industry. Each user manages a firm which designs and markets drugs, such as the “Fountain of Youth” drugs illustrated in the section “Design Objects with Multiple Values.” An MTS could let each user see the all the drug designs that are sent to the market. In contrast, each drug can be marketed under a trade name. A user knows the formulas of drugs produced by his own company, but he must reverse engineer drugs produced by competitors.
Reverse engineering can be simulated in two ways. First, and most simply, a firm can pay a price to have a drug's design revealed. Second, a user can buy filters. Filters are objects that tell whether a design is within a particular object category. By purchasing and testing drugs with various filters, a user can determine the formula for a drug. For example, a user can test a drug with the filter (A####). A positive response means that the drug has an “A” expressed by the first attribute. The firm could further test the drug with the filter (AB###). After multiple attempts the user can learn the formula.
R&D Strategy: All of the aforementioned approaches may influence a user's strategy in developing, or searching for, improved object designs. An important subset of additional influences includes qualities of a business simulation that occur outside of a user's R&D. These include actions, including R&D, of competitors; the user's capability to appropriate the economic value of designs through complementary assets such as distribution channels; increases/decreases in market demand; changes in consumer tastes that favor some technical designs over others; and changes in the prices of raw materials, labor, or capital that might make some designs more or less costly to produce.
Of course, other possibilities exist, but in all cases the richness added allows the user to develop an R&D strategy for effective and efficient R&D. If raw materials costs increase, the user might search for designs that minimize the need for that material. If one technical/market effect is rare but essential, an efficient strategy might test designs first for that effect. In another variation, the attributes and characteristics available for design might change during the simulation. In other versions, the SVF itself might change during the simulation. In fact, this method of modeling R&D can use all of the designs and methods described previously, including those presented in the sections titled “Value Functions” and “Design Objects.”
Unquestionably, the method above allows for both simple and sophisticated models for simulating the R&D. In fact, one could use this method to model other aspects of business, such as modeling market research. Generally, one can use this method for modeling research into the properties and applications of objects.
From the foregoing, it can be appreciated that a computer implemented simulation is described in which there is a set of objects represented by an attribute-characteristic representation and the act of a firm performing research and development is simulated. The simulation includes the steps of accepting one or more search requests from a user, the search request defined with a object category or object design (which is actually an object category that contains one object); evaluating at least one object in said set with a value function selected from the group consisting of VFMOs, MVFs, IVFs, and IAVFs; and selectively providing the user with information concerning said evaluation. One or more of these steps can be repeated a number of times, and possibly only a maximum prescribed number of times.
As well, a computer implemented system and a method for simulating the research and development process can comprise the steps of: defining a set of design objects which are represented by an attribute characteristic representation; defining a set of R&D resources with specificities defined as object categories in an attribute-characteristic representation; accepting one or more search requests from a user, defined by an object category of design objects; comparing the search request with the specificity of the R&D resource to affect/adjust researchable object candidates, costs, information produced, and accuracy, at least in part on the basis of the comparison; evaluating at least one design object associated with the request with a value function selected from the group of VFMOs, MVFs, IVFs, IAVFs; and selectively providing the user with information concerning said evaluation. Note: the design objects and R&D resources can have different attribute-characteristic representations. Alternatively, they can use the same attribute-characteristic representation.
While the previous description of modeling R&D offers specific examples, these should not be construed as limitations on the scope of the invention, but rather as an exemplification of some embodiments.
Simulating Technological Advances, or More Generally, Directing the Simulation to Subsets of Design Objects
During the course of a simulation it can be useful to direct the user to one subset of design objects and then to direct the user to a new subset of design objects. “Directing” means that the simulation makes a subset of the design objects superior to other design objects in a way that is important to users' decisions. The subset with the superior qualities is called the enhanced subset of design objects. The most important application of this method is in simulating technological advances, but there are other applications of this method. These include simulating the opening of new markets, the development of new product applications, the depletion of raw materials, and changes in government regulations.
To see the effect of this method, consider R&D. The simulation directs the user to a particular subset of the design objects. The user performs R&D on these objects and learns combinations of characteristics that create valuable designs. Within this subset he can efficiently create good designs in response to the simulation. Later in the simulation the simulation announces a technological advance and directs the user to a new subset of design objects. The topography of the SVF for this new subset differs substantially from that of the initial subset. The user's knowledge of characteristic combinations is useless, and he must conduct additional R&D to develop new knowledge.
Before introducing specific techniques for directing a user to a subset of design objects, consider a general procedure. The term directing conditions refers to qualities of the simulation that make a subset of the design objects superior to other design objects in a way that is relevant to users' decisions. This superior subset of design objects is called the enhanced subset. A simulation can have a directing condition manipulator to manipulate the directing conditions and thereby adjust the enhanced subset. In this sense the directing condition manipulator “directs” users to different subsets of design objects. To illustrate, the directing condition manipulator creates an enhanced subset. When users make decisions, for example design objects, they will focus upon the enhanced subset. The directing condition manipulator can then alter the enhanced subset, thereby causing users to focus on a new subset. In this sense the directing condition manipulator is “directing” users to specific subsets of design objects.
The directing condition manipulator is similar to the design restrainer depicted in
The preceding presentation did not describe directing conditions or the manipulation of directing conditions. There can be many means of fulfilling these functions. The following are descriptions of three particularly useful methods: (1) manipulating the set of valid object designs, (2) manipulating parameters, (3) and manipulating segments of design objects.
Manipulating the Set of Valid Object Designs: The previous sections “An Illustrative Embodiment: Competitive Search of a Product Value Function” and “Ramifications” described a method of simulating technological advances. From these descriptions, it should be appreciated that a result of changing the domain of one or more attributes is that the set of valid designs for the objects is altered. The set of valid designs is the enhanced subset. The user is directed to these designs because other designs are invalid. Note that when all designs are valid the enhanced set includes all object designs. The directing condition is the quality of the design, valid or invalid. The directing condition manipulator is a mechanism for changing the set of valid designs during a simulation.
If the domain of a particular attribute is expanded to include additional characteristics, then the set valid designs, the enhanced set, is expanded. As a specific example, a manufacturing plant may have had resources to paint cars blue, red or white. As a result of a new source of paint, however, the plant can now also produce cars that are yellow. If the cars at the plant were defined by attributes, one for transmission (standard or automatic), another for air conditioning (yes or no), and one for color (red, blue or white) then the product space would be increased from 12 possible product designs to a total of 16 possible product designs by the introduction of the new color.
It should be understood that the set of valid designs can also be changed by increasing or decreasing the number of attributes in an attribute characteristic representation of the object. For example, the innovation of air conditioning could be added as a new attribute to an existing product line. Thus, in the preceding example, if cars of various colors and one of two transmission types comprised the set of valid designs, the introduction of the new attribute (air conditioning or no air conditioning) would double the set of valid designs.
The set of valid designs can also be restricted by requiring that a combination of characteristics always appear together. For example, a simulation might require that if the characteristic “A” is expressed in a design, then the characteristic “B” must also be expressed. A simulation could require that when a specific attribute expresses “A” another attribute must express “B.” For example, all automobiles that have automatic transmissions must also have power steering. Alternatively, the simulation can require that a combination of characteristics never appear. For example, a simulation could require that characteristics “A” and “B” never appear in the same design. It could require that if a particular attribute expresses “A” that a different attribute cannot express “B.” During a learning session, the simulation can removed or add such restrictions, thereby changing the set of valid designs.
The effect of any of these changes is to represent changes in design opportunities in the simulation. Just as the number of attributes or set of characteristics for a given attribute can be increased, they can likewise be constrained or decreased. Similarly, instead of removing restrictions on combinations of characteristics, a simulation can add restrictions. These changes decrease the set of valid designs, and they can simulate the effect of government regulation, a shortage of supplies, a natural disaster, and the like.
Such changes can be automatically initiated by the central server. (See Section 10, “Hardware Configurations,” below.) For example, the domain of one or more attributes or the number of attributes in the attribute-characteristic representation of the object can be changed automatically in response to a determination of the amount of innovation in the designs being submitted by participants in the simulation. Thus, if the designs that are being submitted to the system for processing by the SVF are approaching maximum values, or not changing substantially from round-to-round, then the system can automatically increase either the number of attributes that can be used to define the objects or introduce new characteristic possibilities that the participants can use to improve their respective designs.
In the event that the domains or attributes are changed, the value of each valid design in the product space preferably remains unchanged. The value of each design would be affected, however, if the SVF were changed or if a characteristic expressed by a system-set attribute were changed.
Manipulating Parameters: Another method of directing the user to a subset of design objects is to provide parameters and/or qualities that impede a user from succeeding with one or more design objects. The unimpeded design objects are the enhanced subset. The parameters and qualities that impede success are the directing conditions. The mechanism that adjusts these parameters and/or qualities is the directing condition manipulator. By adjusting parameters, a directing condition manipulator can direct users away from the impeded design objects and to the enhanced subset. The different between this and the previous method is that the previous method makes some designs invalid. In this method all designs are valid. (Actually, one can consider the previous method an extreme case of this method. In the previous method the impeding conditions are so severe that some designs become invalid.) To illustrate this method, suppose that users design products that have five attributes. Each attribute expresses a letter “A” through “E.” Suppose that at the start of a simulation each characteristic in an object design increases the manufacturing cost according to the schedule in
During a simulation, the simulation can reduce the cost of characteristics “D” and “E.”
The hindering condition need not be cost. It can be any parameter or quality that directs a user to a subset of the design objects. Here are some examples:
Those versed in the art will also realize one could institute this method with a boosting condition instead of a hindering condition, but that this is merely the same method.
Manipulating Segments of Design Objects: One can also direct users to a subset of the design objects by using segmented designs. Consider design objects that have system-set attributes. The system-set attributes constitute the directing condition, and the directing condition manipulator directs the user to a subset of object designs by changing the characteristics expressed by the system-set attributes. The enhanced subset is the set of design objects that have the system-set attributes in their current settings. By changing these attributes the directing condition manipulator directs the user to a new subset of design objects. After such a change the user “operates” in a different portion of the value function. This is true even if the user does not know of the system-set attributes. For example, suppose that a simulation has design objects that have five design attributes and two system-set attributes. Suppose at the start of a simulation the system-set attributes both express the characteristic “A.” At the start of the simulation the enhanced subset is (#####AA). During the simulation the directing condition manipulator changes one of the system-set attributes to “B.” The new enhanced subset is (#####AB).
Segmented designs provide other mechanisms for directing users to subsets of objects. As an illustrative example, consider pharmaceutical R&D. As described above, in the subsection titled “Examples of Using Segment Objects,” which is in the section titled “Design Objects,” one can simulate pharmaceutical R&D with a design object having two segments. One segment represents the drug being designed and the other represents the target. During a simulation a directing condition manipulator can add and/or removed targets. The addition of targets simulates scientific advances such as discoveries arising from genetics. Also, the directing condition manipulator can favor some targets over others by boosting the value of designs that use some targets and decreasing the value of designs using other targets. Generally, by adding and removing targets and by adjusting qualities of the simulation in response to targets, a directing condition manipulator can direct users to subsets of the design objects. Notice that this method is similar to two of the methods already described. Adding and removing targets is the similar to changing the set of valid designs. Adjusting the value of designs according to their target is similar to having a hindering or boosting condition. The use of segmented designs enhances these methods.
To further illustrate the usefulness of segmented designs to altering the enhanced subset, suppose that design objects represent products and product applications, as described in the subsection titled “Product Applications,” in the section titled “Design Objects.” One segment represents product designs and the other represents applications of products. The simulation can define market segments by having a finite number of applications, where each application is a different market segment. The simulation can direct users to subsets of design objects by changing the characteristic expressed by the attributes representing market segments. This represents a change in the application. For example, suppose the market for adhesives is described with a system-set attribute. This attribute represents the surface to which adhesives are applied. It can be “A” or “B” representing metal and plastic. Suppose that at the start of a simulation all customers use adhesives for binding metal. During the simulation some customers start using adhesives to bind plastic, thereby opening a new market segment. The simulation would affect this by changing the system set attributes for some customers from “A” to “B.” Alternatively, both applications can exist at the start of a simulation. During the simulation the market segment for binding metal can shrink while the market segment for binding plastic grows.
Three Methods: The examples above illustrate three methods, or three types of directing conditions. First, the directing condition is a set of rules or calculations involving attributes and characteristics from the attribute-characteristic representation of objects. The directing condition manipulator adjusts these rules and/or calculations, and the enhanced set is determined by applying the rules and/or calculations. Second, the directing condition is the value function itself. The directing condition manipulator alters the value function, and the enhanced set is given by the value function topography. To illustrate with the mountain range metaphor, the enhanced set can be a set of designs sitting beneath a high, broad plateau. Third, the directing condition is the characteristics expressed by system-set attributes. The directing condition manipulator changes the enhanced set by changing the characteristics expressed by the system-set attributes. For example, the enhanced set can be the object category defined by the current setting of the system-set attributes. All other designs are impossible to create because the simulation controls these attributes. The system-set attribute method is based upon attributes and characteristics, so it can be considered an example of the first method. However, because it is particularly useful, one might consider this a distinct method.
MTS Dynamics Can Also Direct Users: The normal processes of an MTS can also direct users to a subset of design objects. The following example illustrates. Consider an MTS where users design and manufacture computers. Design objects represent computers. Attributes represent components, such as the mother board, microprocessor, and hard drive. Characteristics represent types of components, such as different types of microprocessors and hard drives. To manufacture computers users must purchase the components at prices given by demand curves. If users increase demand for a particular type of microprocessor, the price increases. This directs users to the subset of designs objects that lack this type of microprocessor.
As another example, consider a simulation where users design and sell glue. Suppose that glue is sold in two markets, industrial and consumer. Suppose that competition in the industrial market increases and as a result prices drop. Users might shift their design efforts to exploit the still profitable consumer market.
If normal MTS dynamics can direct users to subsets of design objects, why use a directing condition manipulator? The manipulator gives an MTS control over the dynamics of a simulation. Generally, it “reverses” the direction of causation. Instead of simulation dynamics directing users, a directing condition manipulator directs the users and thereby affects simulation dynamics. With a directing condition manipulator a simulation can cause users to face situations that are particularly important for practice and learning.
This illustrative embodiment serves three purposes. First, it illustrates another application of the present invention. Specifically, it presents an MTS for leaming to manage a product development portfolio. Second, it illustrates different types of decisions that the user can make. The previous descriptions show users inputting design objects and/or object categories. This new illustrative embodiment illustrates this arrangement, but it also illustrates another arrangement. It shows a system where the users' decisions do not include submitting any design objects or object categories. Third, this illustrative embodiment shows additional uses of the present invention. It can be used to diagnose and/or evaluate business decisions, policies, and strategies. This illustrative embodiment is simple. It does not illustrate many of the methods previously described, such as a directing condition manipulator. It does not simulate manufacturing and uses the most simplistic model of a market. Those versed in the art can augment this illustrative embodiment with the methods previously described, simulations of manufacturing and complex markets, and/or other modifications.
Product Development Portfolio: A product development portfolio is a means of organizing product development. It helps a business to decrease the time needed to create new products, increase the number of new products a company creates, coordinate product development with business strategy, and manage risk. For an introduction to portfolio management, see chapter 1 of Portfolio Management for New Products by Cooper, Edgett, and Klienschmidt. Briefly, the product development portfolio contains product development projects. Projects can be anything from basic research in new science to modifying an existing product.
The job of portfolio management involves the following tasks: (1) keeping the portfolio full of high value projects, (2) allocating resources among projects, and (3) canceling and promoting projects. These tasks are preformed to meet market opportunities, keep pace with technical advance, and maximize the return on R&D investments. Managing the portfolio is difficult because each task affects risk, uncertainty, the probability of successfully creating new products, and the time required to create new products. Consider, for example, the following common mistake. New projects are added to the portfolio faster than existing projects are cancelled. As the number of projects in the portfolio increases, the portfolio becomes bloated. Resources are spread thinly and product development slows to a crawl. Capital must be committed for longer periods of time, increasing risk and capital costs. To get products through stage 3, managers respond to this problem by allocating more resources to projects producing incremental innovations. These projects can be finished quickly, but they produce little technical advance. Because of these errors, the product development portfolio requires too much capital, has a low rate of return, and is technologically uncompetitive.
The illustrative embodiment is an MTS for leaming to manage a product development portfolio. Its description begins by describing design objects, the SVF and the user's goal. The illustrative embodiment's components are described next. When all of the components have been described, their integration and the operation of the illustrated embodiment are discussed. Finally, modifications and alternatives are described.
Design Objects and SVF: In this illustrative embodiment design objects are represented with an attribute-characteristic representation having two segments. One segment has seven design attributes and the other segment has two attributes representing the stage of R&D. Each design attribute can express one of eleven characteristics, represented by the letters A-K. Each R&D stage attribute can express one of two characteristics, represented by the numbers 0 and 1. The following are examples of object designs: (ABC DE FG 00), (DDG JH BB 11), and (KFD GE BA 10). However, where appropriate, this description will ignore the R&D stage attributes and present designs as in the following examples (ABC DE FG), (DDG JH BB), and (KFD GE BA). R&D stage attributes are described separately below. (Spaces between attributes make reading designs easier.)
The SVF for this illustrative embodiment is an nk-landscape function, and the configuration described here is the presently preferred configuration. Potentially, one could find a better configuration or even a better SVF. As previously described, the nk-landscape function has four important parameters: (1) the number of attributes, (2) the number of characteristics, (3) the average number of interactions per attribute, (4) and the arrangement of interactions over attributes. The attributes and characteristics have just been described, so consider the interactions. There are two types: (1) intrasegment interactions among the design attributes and (2) intersegment interactions through which the R&D stage attributes interact with the design attributes. These intersegment interactions are described below, so consider the intrasegment interactions. There should be enough intrasegment interaction among the design attributes so that the average number of interactions per design attribute is between 2 and 3. Since there are seven design attributes, there should be 14 to 21 interactions. Note that a two-way interaction counts as two interactions, and a one-way interaction counts as one interaction. The nk-landscape function can use a symmetrical or asymmetrical distribution of intrasegment interactions among design attributes. An asymmetrical distribution has consequences for R&D, described below. The SVF evaluates only the portion of the design object consisting of the design attributes, although the R&D stage attributes affect these evaluations. Finally, the SVF assigns designs a value from 0 to 10. For an nk-landscape with seven design attributes, this means that each attribute can contribute a value between 0 and 1.429 to a design's value. (For a discussion of nk-landscapes, one-way and two-way interactions, the distribution of interactions, and evaluating a portion of a design see Section 6, subsections titled “The nk-Landscape Function” and “Evaluating a Portion of a Design.”)
The User's Goal: In the illustrative embodiment the user manages a product development portfolio of a simulated company. He designs products to fulfill a goal. The goal can be simple. For example, the goal can be to create a product with a value greater than a particular value. Alternatively, the goal can be to maximize the value of the best product designed within a limited budget, a set number of rounds, or a limited amount of time. In these configurations the user's success is measured by the value of his best object design. Simulations with these goals lack a simulated marketplace. This is noteworthy: the portfolio management MTS need not simulate a marketplace. Alternatively, a product development portfolio MTS can simulate a market. The user must then design products to exploit market opportunities. Given the techniques described herein (particularly in Sections 2, 7, and 8) and the previous art of modeling markets, one can develop sophisticated market models. The illustrative embodiment presented here uses a simple model of a market. This allows the user to focus on managing the product development portfolio.
The illustrative embodiment models markets as a series of market opportunities. Each opportunity has four characteristics: an identifying number, a required value, a payoff, and an expiration date. To exploit an opportunity the user must create a design that exceeds the required value before the expiration date. If the user accomplishes this task, he receives the payoff from that opportunity. For example, suppose there is an opportunity with a required value of 8.45, a payoff of $0.5 million, and expiration date of round 7. (Playing the simulation in rounds is described below). If the user can create a design with a value exceeding 8.45 by the end of round 7, the simulation credits his score with $0.5 of revenue. If the user cannot design such a product by the end of round 7, the opportunity expires and is lost. Because opportunities have required values and expiration dates, they present users with design goals and “windows of opportunity.”
The required values should be high. To understand why, consider a random search for good designs. If a require value is higher than the value of only 30% of the object designs, finding a design that exceeds the required value is easy. Random search should succeed 70% of the time. Now suppose that a required value is higher than the values of exactly 95% of the object designs. Random search will succeed, on average, only 5% of the time. To make the simulation difficult, required values must exceed the values of a significant percent of designs. To help set required values, one can randomly sample from the value function and create a histogram of the design values. From the histogram one can estimate values that exceed any percentage of design values. In technical terms, one uses the histogram to estimate the percentile rank of a value. By raising or lowering the percentile rank of required values one makes the simulation more difficult or easier.
The simulation presents the user with a table of the simulation's market opportunities. The table is displayed on a form that the user can view throughout the simulation.
Projects and Portfolios: Generally, the user's goal is to maximize his simulated firm's revenues while minimizing the cost of designing products. He should maximize the return on investment from product design. To do this he must create designs with high values, and he must do so before opportunities expire. Section 3 of this specification, titled “Ramifications,” describes how users can create high value designs. Specifically, its subsections titled “Designing Products” and “Projects” explain the use of a portfolio of projects. Summarizing these subsections, the large number of designs and the interactions among attributes make designing objects difficult. The user can be more effective at designing high value objects if he focuses on a small set of design objects that contains high value designs. To accomplish this, the user proposes and evaluates an object category. Recall that a user evaluates an object category by evaluating several designs from the category. If these designs have high values the category might contain many high value designs. If this is the case the user focuses his efforts on designs in that category. If the category is not promising, the user proposes and evaluates another object category.
To illustrate, consider the illustrative embodiment. There are seven design attributes, and each one can express one of eleven characteristics, so there are nearly 19.5 million designs. Because of interactions between design attributes, finding the best of these designs is a difficult problem. To ease this task the user proposes and evaluates a category. For example, the object category (A## #C ##) contains approximately 160,000 objects. If evaluation suggests that the category contains high value designs, the user's task becomes much simpler. He need only search among 160,000 choices, and many of these are good designs. This method of simplifying search is given a name. An object category used to focus the design task is called a project. Using this name, the user proposes and evaluates projects. Projects that evaluate poorly are cancelled. Promising projects are retained.
Suppose a person considers one project at a time. Evaluating the category by evaluating only a few designs from the category creates uncertainty. The situation is analogous to the statistical process of population sampling, where a small sample is taken from a large population. The user can reduce uncertainty by testing more designs, but this is inefficient. The reason for using projects is to discover where to apply effort before expending considerable resources. The dilemma is resolved by using several projects simultaneously. The user can invest a small amount of resources in evaluating several projects. Using several projects decreases risk, because the user is not “putting his eggs in one basket.” The user can then direct larger investments as the initial evaluations indicate. By dynamically allocating investment in response to information, the user can manage both risk and investment to maximize the effectiveness of product development. In sum, the user can start with several projects and dynamically allocate resources as the task of designing objects produces information. The user can both manage risk and efficiently design high value objects. The collection of projects is called a portfolio, illustrated in Section 3 by
A Three stage Portfolio: The portfolio illustrated by
The illustrative embodiment uses a three stage portfolio, similar to that just described. The portfolio is illustrated with three tables. These tables are presented to the user with a form that is available throughout the simulation.
By looking at each figure one can understand how the three stages work together. To design objects, the user starts by conducting R&D (described below) to create initial hypothesizes of product categories that contain high value products. The user is instructed to hypothesize object categories defined on two attributes. These hypothesizes become stage 1 projects in the portfolio.
Research and Development (R&D): The descriptions above refer repeatedly to R&D. How does the illustrative embodiment simulate R&D? It uses a simple version of the method described above in Section 8, subsection titled “Modeling R&D.” Basically, a user submits designs to the simulation. The simulation evaluates the designs and records the results. In this illustrative embodiment the user must follow the following procedure. The user must first specify a project or subproject in which to conduct R&D. The user can then submit a design that is contained in the project's definition. Recall that projects are defined by object categories. A design is “contained in” a project if it is a member of the subset of design objects specified by the object category. For example, (A## J# ##) is the definition of project 11 (see
The Test Designs area of the R&D Laboratory is where the user tests designs. The user can only test designs that are contained in the selected project's definition. The Project Definition list box (6206) displays this definition as a reminder to the user. After the user has selected a project from the Select Project combo box, and in the case of a stage 2 project subsequently selecting a definition from the Project Definitions list box, the user can test designs. To test designs, the user enters the designs to be tested into the textboxes 6210, 6212, 6214, 6216, and 6218. The five textboxes enable the user to test up to five designs at once, although the user can also test fewer designs. After entering the designs the user clicks the Test Designs button (6220). The simulation checks each design to ensure that it is contained within the project's (or subproject's) definition. Acceptable designs are then evaluated with the SVF, and the results are stored by the simulation. After the test, the textboxes are cleared so that the user can perform more tests. The “# of Tests Remaining” is updated as well (6208).
How does the user decide which designs to test? The user reviews the results from a project's previous R&D tests, and he tries to identify combinations of characteristics that produce high value designs. The user then tests designs to assess his hypotheses. For example, suppose that the user is conducting R&D on subproject 10.1. Its definition is (KBB K# ##). After reviewing the results of project 10's previous R&D tests, the user might hypothesize that designs that end with “A” have high values. For subproject 10.1, this hypothesis is that designs in the object category (KBB B# #A) have high values. The user can assess this hypothesis by testing the designs in that category, for example (KBB KD AA), and (KBB KC EA), (KBB KD JA). If these designs have high values, the hypothesis might be correct. If the designs have low values, the user should review the R&D data again and develop a new hypothesis.
How does the user review the data from previous R&D tests? The remaining portions of the R&D Laboratory are for this purpose. These are the View R&D Data area (6222), the R&D Test Results table (6224), and the R&D Test Results graph (6226). First, consider the R&D Test Results table (6224). When the user selects a project from the Select Project combo box (6202) the results of all the project's R&D tests are displayed in the R&D Test Results table. If a stage 2 project is selected, all tests done for the project and all of its subprojects are displayed. Each row of the table presents the result of a test. The table has five columns. The first column shows the project or subproject in which the test was conducted. The second column lists the test number. The tests within a project are numbered so that the user can identify them easily. Tests are numbered in the order in which they were conducted, starting with the first test made when the project was in stage 1. Column 3 shows the stage that the project was in when the R&D test was conducted. Column 4 displays the designs that were tested, and column 5 displays the design values revealed by the tests. Now consider the R&D Test Results graph (6226). The graph shows the value of the tests listed in the table. The graph displays the data in the order that is presented in the table, with the first test in the table being at the extreme left of the graph and the last test in the table being on the extreme right of the graph.
Identifying good combinations of characteristics from a long list of data is difficult, so the View R&D Data area provides tools to filter and sort the data displayed in the R&D Test Results table and graph. The user can filter the data in three ways: by value, by object category, and by stage. To filter by value the user uses the area titled Filter By Value (6228). In this area there is a greater-than sign and a less-than sign. Next to each sign is a textbox for input (6230 and 6232). The user puts values into these fields and then clicks on the Filter By Value button (6234). For example, the user can put a 7 in the greater-than textbox (6230) and a 2 in the less-than textbox (6232). After pressing the Filter by Value button (6234), the R&D Test Results table (6224) displays only the tests of designs that have values greater than 7 or less than 2. If the user desires he can use only the greater-than criteria or only the less-than criteria. The user simply leaves one of the textboxes blank. For example, to filter for designs with values greater than 7, the user would leave the less-than textbox blank.
To filter R&D tests by object category the user uses the Filter By Category area (6236). There is a textbox in which the user types in an object category (6238). The user then clicks on the Filter By Category button (6240). The R&D Test Results table (6224) will now show only the tests of designs that are contained in the designated object category.
The user can also filter R&D data by stage. The user uses the Filter By Stage area (6242). He selects the stage from the combo box (6244). (The combo box is configured as a dropdown list box.) The user then clicks on the Filter By Stage button (6246), and the R&D Tests Results table and graph will display only data from the selected stage.
In addition to the filters, a user can sort the R&D test results that are displayed in the R&D Test Results table. The user uses the Sort area of the R&D Laboratory (6250). He can sort data by the project number, test number, or design value. The user selects the type of sort from a list displayed by a list box (6252). The user then determines whether the sort should be ascending or descending by selecting a choice from the option buttons (6254). The user then clicks the Sort Project Data button (6256) and the test results currently displayed in the R&D Tests Results table (6224) are sorted as specified. The R&D Tests Results graph (6226) is similarly affected. The example in
To further analyze the R&D tests results, the Statistics area (6258) displays statistical information about the R&D results being displayed in the R&D Test Results table (6224). The Statistics area shows the maximum value (6260) and the average value (6262) of tests displayed in the R&D Test Results table.
In addition to conducting tests and viewing test data, the R&D Laboratory enables a user to add new projects to the portfolio. Only two types of projects can be added: stage 1 projects and stage 2 subprojects. First, consider stage 2 subprojects. Recall that a stage 2 subproject is defined by an object category that (1) is defined on exactly four attributes and (2) is contained in the object category that defines the project. To create a stage 2 subproject, a user first selects a stage 2 project to conduct R&D upon. He selects the project with the Select Project combo box (6202), as described above. After selecting a stage 2 project, the user sorts and filters the project's R&D data, and he tests designs to create new R&D data. The user performs these actions to identify a good object category for a subproject. When the user believes he has found a subproject definition that contains high value designs, he types it into the New Project Definition textbox (6264). He then clicks the Add New Project button (6266). The simulation checks to ensure that the subproject is defined on four attributes and is contained in the definition of the selected project. If this is the case, the subproject is added to the project. The new subproject is given a number and added to the Project Definitions list box (6206). The user can now conduct R&D for the new subproject.
Now consider the task of creating new stage 1 projects, illustrated by
To create the new project, the user filters and sorts the R&D test results searching for combinations of characteristics that produce high value designs. Specifically, he tries to identify an object category defined on exactly two attributes that contains high value designs. To aid this effort, the user can produce additional data by conducting R&D tests, using the Test Designs area. Because the user is not doing R&D within a project, but is doing exploratory research, the R&D tests are not constrained by an object category. Accordingly, the Project Definitions list box (6450) displays (### ## ##). When the user has identified a suitable object category, he types the category into the New Project Definition textbox (6460). In the example of
Why Three Stages of R&D? After seeing how R&D works, one can ask, “Why not conduct all R&D in a single stage?” The three stages are necessary because of two qualities of R&D. Before describing these qualities, the following terminology is defined for convenience. Testing a design in efforts to create a stage 1 project is called stage 0 R&D. Testing a design for a stage 1 project is called stage 1 R&D. Testing a design for a stage 2 project or subproject is called stage 2 R&D, and testing a design for a stage 3 project is called stage 3 R&D. Now consider the two qualities of R&D that give rise to stages. The first is cost. Testing a design is costly and varies with the stage of a project. Stage 0 R&D and Stage 1 R&D cost $1,000 for each design tested. Stage 2 R&D costs $5,000 for each design tested. Stage 3 R&D costs $25,000 for each design tested. The second quality is reliability. The reliability of R&D varies with each stage. Stage 1 R&D is not very accurate. There will be significant differences between the design values reported by stage 1 R&D and the true value of a design. Stage 2 R&D is more reliable, but still inaccurate. Stage 3 R&D is completely accurate, meaning that tests done on stage 3 projects will reveal the correct values of designs. Given this situation, the most efficient means of designing objects is to use the inexpensive stage 1 R&D to identify a large category of promising object designs. Stage 2 R&D is then used to narrow the category further. When the costly, reliable stage 3 R&D is performed, the user has narrowed the possibilities to a small set of promising candidates. This way, stage 3 R&D is not wasted on unpromising designs.
The simulation controls the accuracy of R&D using the R&D stage attributes. The R&D stage attributes interact with the design attributes as shown in
The R&D stage attributes affect the accuracy of R&D. The simulation defines the true value of a design as the value given when both R&D stage attributes are set to one. Stage 3 R&D gives the true value. Because S2 interacts with two design attributes, stage 2 R&D gives inaccurate results. Stage 1 is even less accurate. Both R&D stage attributes are changed (from one to zero), and together they interact with four design attributes. The accuracy can be measured exactly with correlation. One can correlate the values assigned by stage 2 R&D with the values assigned by stage 3 R&D, and this is a measure of the accuracy of stage 2 R&D. One can also correlate the values assigned by stage 1 R&D with those assigned by stage 3 R&D. The calculations will show that correlation between stage 2 and stage 3 R&D is greater than the correlation between stage 1 and stage 3 R&D, so stage 2 R&D is more accurate.
While cost and reliability necessitate a staged model, the distribution of intrasegment interactions among the design attributes can accentuate the importance of the stage process. When interactions are distributed asymmetrically, the attributes affecting the most other attributes are called highly interactive attributes. They are critically important because changing the characteristic expressed by a highly interactive attribute greatly changes a design's value. Therefore, a user should use stage 0 R&D to discover which attributes are highly interactive and to identify good characteristics for these attributes to express. Stage 1 R&D should test the results. When a good project definition is found, stage 2 R&D and stage 3 R&D are used to “fill-in” the characteristics of the remaining attributes. In this way product development exploits the asymmetrical distribution of interactions.
Promoting Projects, Canceling Projects, and Allocating Resources: The presentation above describes the three stages of the portfolio and how the user conducts R&D. To manage a product development portfolio the user must allocate capital to finance R&D. The user must also promote and cancel projects. The user accomplishes these tasks with the Project Decision form, as depicted in
The form also has a table for inputting decisions (6630). The table has one row for each project in the portfolio, plus a row that represents the task of conducting R&D to create new stage 1 projects. The first column of the table gives the projects' numbers. The second column gives the stage of the projects. The third and forth columns are where the user inputs decisions. The third column is titled “Status,” and this is where the user inputs the status of each project. The user can affirm a project's status as active, or he can tell the simulation to cancel or promote a project. For stage 2 projects this column is split into two parts. If a stage 2 project is promoted, the user must tell the simulation which subproject to promote. The second part of the column is used to input this decision. The fourth column is titled “Budget.” This is where the user inputs his resource allocation decisions. Because it is unique, consider the row of the table that represents the task of conducting R&D to create new stage 1 projects. The project's number is listed as “New Projects.” Because this “project” is R&D for creating stage 1 projects, its stage is listed as zero. Because the user always has the option to fund the development of new stage 1 projects, its status is always “Active.” This cannot be changed, however if desired, the user can allocate this task a budget of zero.
The description above shows how the user cancels and promotes projects. There are two circumstances in which the simulation, rather than the user, promotes and cancels projects. When either situation occurs, the simulation notifies the user with a message box. The first situation occurs when a market opportunity expires. The “window of opportunity” has closed, and all projects that have the opportunity as their goal are cancelled and removed from the portfolio. Second, if stage 3 R&D identifies a design that satisfies the project's goal, the project is successful. The simulation immediately promotes the project out of the portfolio, and the project's market opportunity is satisfied. In addition, all other projects that had this market opportunity as their goal are cancelled. A market opportunity can only be satisfied once. The remaining budgets of successful and cancelled projects are removed from the simulation, and the user is credited with the payoff from satisfying the market opportunity.
Note that the user must test the “winning” design with stage 3 R&D. Testing a “winning” design with stage 1 or stage 2 R&D is not enough. For example, suppose that the user has a stage 1 project, and he tests the design (ABC DE FG). Suppose that the design's true value exceeds the project's goal. Though the project was tested with stage 1 R&D, the project had not achieved its goal. Suppose that the user promotes the project to stage 2 and again tests the design (ABC DE FG). The project has still not satisfied its goal. Now suppose that the user promotes the project to stage 3, and again he tests the design (ABC DE FG). This time the project achieves its goal. The project is promoted out of the portfolio, and the user receives the payoff.
Project Evaluation: To help the user make promotion, cancellation, and resource allocation decisions, the simulation presents the user with a form containing a Project Statistics table, as illustrated in
To illustrate the data presented in Columns 1-7 consider project 7, depicted in row 3 of
Columns 8 and 9 of
Though the Project Statistics table is useful for making decisions, the simulation provides another table to help the user. With this table the user scores and ranks projects. Scoring and ranking is a popular method of evaluating projects when managing a product development portfolio. Managers decide upon a set of criteria on which to measure projects. Each project is then scored on each criterion. A project's scores are then combined, usually as a weighted sum, to create a total score for the project. The projects are then ranked by their total scores.
The illustrative embodiment enables the user to perform this process with a Project Evaluation form (7010), illustrated in
The other rows of the table are for projects. Notice that each project, and its stage, is listed on the left side of the table (7060). In each project's row, in columns for the criteria, the user scores each project on a three point scale. A project gets 1 point if it measures poorly on a criterion. It gets 2 points if it is measures adequately on a criterion, and it gets 3 points if it is measures outstandingly on a criterion. The user is free to define poorly, adequate, and outstandingly any way he chooses. As an example, consider scoring projects for the payoff criterion. A project pursing a market opportunity with a payoff of less than $1 million can get 1 point. A project pursing a market opportunity with a payoff between $1 and $1.3 million can get 2 points, and a project pursing a payoff exceeding $1.3 million can get three points. For some of the criteria the user must consider a project's stage when scoring the project. This is because the stage significantly impacts how a project measures on some criteria. An example is the additional investment criteria. The additional investment needed by a stage 1 project is more than that needed by a stage 3 project. The stage 1 project must have enough funds for remaining stage 1 R&D plus funds to pay for subsequent stage 2 and stage 3 R&D. To score projects on the six criteria, the user will find the Project Statistics table (
When all of the scores are assigned, the user clicks the Total Scores button (7030). The form calculates the total score for each project and displays the scores in the Total Score column (7070). In addition, the simulation ranks the projects by the scores. The project with the highest score gets a rank of one. The project with the second highest score gets a rank of two, and so on. The project rankings are displayed in the Project Rank column (7080). Total score calculation is a weighted sum. For each project, the simulation calculates its score by multiplying each of its criteria scores by its criterion's weight. The resulting weighted scores are added.
The illustrative embodiment provides only two means of evaluating projects: the Projects Statistics table and the Project Scoring form. One versed in the art could devise other means. For example, a simulation can have a user make additional estimates about projects and use these estimates to estimate the net present value (NPV) and internal rate of return (IRR) for each project. The user can then use these financial calculations when making promotion, cancellation, and resource allocation decisions. The simulation could also provide bubble charts that graphically compare projects. For example, one axis of the chart could be payoff, the other axis of the chart could be the probability of success, and the size of the bubble could represent the additional investment needed. In fact, one versed in the art could design project evaluation means as complex and/or realistic as desired.
Performance Evaluation: Simulations of a product development portfolio can record decisions and simulation events during a training experience. A simulation can analyze this information and provide feedback to help the user learn. Those versed in the art can conceive of many types of analysis, especially for simulations that have realistic models of markets. This illustrative embodiment provides only a simple analysis. The embodiment records cash flows, specifically the R&D investments and revenues received each round for each market. The simulation also records project history, including each project's goal, stages, and outcome for each round of the simulation. When the simulation ends, it analyzes this data and presents the results to the user. The user can judge his success from this information.
At the end of a simulation, the simulation presents the user with two charts and a table.
In addition to the charts the simulation also presents the user with a table.
Architecture: The illustrative embodiment must store and display data about market opportunities, projects, and R&D tests. This data is simple and can be stored in tables. Furthermore, to fill the charts the simulation needs only a few types of queries to modify and read the data tables. For this reason, the illustrative embodiment could be programmed as a spreadsheet application, for example using Excel and Visual Basic for Applications. The program's architecture would be built around several tables of data. Other architectures are possible. For example, projects, market opportunities, and R&D tests could be objects. One of the properties of the project object would be a list of R&D tests conducted for that project. Another property of the project object would be a reference to a market opportunity object. In this way one can create the program in Visual Basic or another suitable programming language. Whether one uses a spreadsheet model or an object oriented programming model, the techniques needed to design and code the simulation are well known to those versed in the art.
Operation: The description above presents the R&D Laboratory and forms and tables for users to view the projects' data, evaluate projects, input decisions, and receive analysis and feedback. The following description shows how these components operate together during a simulation.
While these steps are integrated, for the sake of clarity,
If all R&D budgets are exhausted, the simulation moves from step 7540 to step 7545. The simulation identifies all market opportunities that will expire at the end of the round and marks them as expired. This has three consequences. Projects pursuing that market opportunity have failed. They are removed from the user's portfolio, and the user is notified with a message box. The simulation also records project data. Second, the table of market opportunities changes to indicate that projects have expired. To illustrate,
The first few rounds of a simulation have particular qualities. They are used to fill the portfolio. At the start of the simulation, round 1, there are no projects in the portfolio. Therefore, the user does not evaluate the projects or the portfolio. Likewise, there are no promotion, cancellation, or resource allocation decisions. Effectively, steps 7510 and 7515 are skipped. The simulation moves from step 7505 to step 7520, where all capital for round 1 is invested in creating new stage 1 projects. Rounds 2 and 3 are also particular. In round 2 the portfolio contains only stage 1 projects, assuming the user created some in round 1. The user conducts R&D on these projects in round 2, and in round 3 he promotes some of them to stage 2. In round 4 the user promotes some stage 2 projects to stage 3. Round 4 is the first round which has projects in all three stages. Because the portfolio is not “full” until round 4, at the earliest, the capital for rounds 1-3 can be reduced. Round 1 can provide the user with enough capital for stage 0 R&D. Round 2 can provide the user with enough capital for stage 0 R&D and stage 1 R&D. Round 3 can provide the user with enough capital for stage 0 R&D, stage 1 R&D, stage 2 R&D. Beginning in round 4 the simulation provides enough capital for R&D in all stages.
Other Measures of Time: In the illustrative embodiment, the simulation progresses in rounds. This is not necessary, for there are other ways to mark time in a simulation. Simulation time can be fixed to real time. For example, one hour of real time can represent one quarter of a year in simulation time. Alternatively, R&D can mark time. For example, conducting 50 tests could require one quarter of a year. After every 50 R&D tests, the simulation advances to the next business quarter. Finally, the simulation need not keep track of time at all. The user can be given one large budget, and the simulation ends when the budget is exhausted. With any of these arrangements the user can view the forms and make decisions in any order. The step-by-step format described above need not be followed. For example, a user can evaluate projects while conducting R&D, and he can make decisions whenever and as frequently as he wishes.
Hidden Attribute-Characteristic Representation and SVF: Thus far, this specification has described MTSs where users design objects and/or object categories and then inputs them into the simulation. This need not be the case. The user can make decisions without ever designing/inputting object designs and object categories. In fact, the user need not be aware of the attribute-characteristic representation or of the value function. These can be internal components of the simulation that are affected by the user's decisions.
To illustrate such an arrangement consider some alterations to the illustrative embodiment. There is no R&D Laboratory, and the user does not conduct R&D. Because there is no R&D Laboratory, the user does not create or input object designs. Furthermore,
The simulation continues with step 7525. In step 7525 the altered embodiment automates R&D. This occurs “inside” the simulation, and the user does not see it. To automate R&D the altered embodiment must conduct R&D tests and create new stage 1 projects and stage 2 subprojects. First, consider the R&D tests. For each project the simulation randomly generates designs and tests them with its SVF. To illustrate, suppose that the user allocates $20,000 to project 11. Project 11 is a stage 1 project, and stage 1 R&D costs $1,000 per test. The computer will randomly generate 20 designs that are contained in project 11's definition and it then it will evaluate these designs with its SVF. Now consider creating new projects and subprojects. To create stage 1 projects the computer checks the budget allocated to “New Projects” in the Project Decision form. For each $10,000 allocated to “New Projects” the simulation will create one new stage 1 project. To create a stage 2 project, the simulation checks the capital allocated to each stage 2 project. For every $45,000 invested in a stage 2 project the simulation will create a subproject. Resources allocated to subprojects are not counted. For example, in
After conducting all R&D the simulation checks stage 3 projects to see if any met their goal (7730). If a project meets its goal, the simulation removes or promotes it out of the portfolio and credits the user with its payoff. The simulation also cancels all other projects that pursued the same market opportunity, and it marks the market opportunity as satisfied. This market opportunity cannot be the goal for any new projects. Finally, the simulation records these events for the post simulation analysis. Note that if two stage 3 projects satisfy the same market opportunity in the same round, the user receives only one payoff.
If there are no successful stage 3 projects, or after successful projects are removed, the simulation checks if there are market opportunities that expire at the end of the current round. If so, the simulation expires these market opportunities, as described above (7745). It then checks if there are any market opportunities that have not expired (7750). If so, the simulation starts a new round at step 7705. Otherwise, the simulation ends by analyzing the recorded data and displaying a performance evaluation to the user (7755).
In this altered embodiment the user does not design/submit objects or object categories. In fact, he never sees an object design or object category, and he can be completely unaware of their existence inside the program. Likewise, the user knows that projects are being evaluated, but he does not know that projects are defined as object categories or that project statistics are based upon evaluating object designs with a SVF. The user can be completely unaware of design objects, the attribute-characteristic representation, and the SVF. These are internal components that the user does not see. The simulation manipulates these components, and the user's decisions affect that manipulation. The altered embodiment is one example of this arrangement, but those skilled in the art can think of others. For example, instead of inputting decisions the user can input or select policies and rules that guide the simulation. Those versed in the art could modify the descriptions and techniques described in Sections 1-11 of this description so that the current invention and its embodiments do not require inputting object designs or object categories.
Diagnosis and Evaluation: MTSs are primarily used for training and learning, but they are used for other purposes too. One such application is diagnosing errors and problems in business decisions and policies. With the MTS, the user tests various policies, strategies, and decisions, and observes the results. Consider the illustrative embodiment. How poorly must a project perform before being cancelled? A user can test different policies. At the end of the simulation, the analysis presented in
Alternative Configurations: The illustrative embodiment presents a basic design. One can modify this design and/or add any of the techniques described above. For example, one can use a sophisticated model of markets and add directing conditions and manufacturing. One can utilize the various techniques described above, including those described in section 6, “Value Functions,” and section 7, “Design Objects.” One can extend the illustrative embodiment to include multiple users who each manage a portfolio and who compete against each other to exploit market opportunities. Furthermore, the projects need not be product development projects, but can be research projects, for example projects exploring new technologies. Generally, the entire illustrated embodiment can be seen as a method for modeling a portfolio of research projects, which is more general than a portfolio of product development projects. The examples presented herein should not be construed as limitations on the scope of the invention, but rather as an exemplification of some embodiments. Accordingly, the scope of the invention should be determined not by the embodiments illustrated, but by the appended claims and their legal equivalents.
Some SVFs need not require a computer. An MTS built from such a SVF need not use a computer. For example, it might use charts or a calculator. This makes possible embodiments of the present invention in business simulation board games. However, most embodiments will use a computer. Without limiting the scope of the present invention, the following three computer configurations will be typical for the current invention: (1) stand alone, (2) multiple computers with floppy disks, and (3) computer network.
Stand Alone: An embodiment of the present invention could run on a single computer. The computer would have a program that includes a simulated business situation and the display. The display might include multiple forms or graphical displays to present information and receive inputs from a user.
Multiple Computers with Portable Memory Devices: Many MTSs use multiple computers and portable memory devices to ferry information between the computers. Usually, the portable memory devices are floppy disks, but other devices, such as flash memory, can be used. In this configuration, there is an MTS administrator and several teams of users, each team making decisions. The administrator has one computer running the simulated business situation. Each team of students has a computer running an interface program that displays results and accepts decisions.
Usually, these simulations are run in rounds, but this need not be the case. Each round begins with the administrator giving each team a floppy disk, or other portable memory device, with information about the simulated business situation. Each of the teams views the information with the interface program on its computer and makes decisions. The interface software then saves the decisions on the floppy disk. The teams then return their floppy disks to the administrator, and the administrator loads the decisions into the simulation running on his computer. The simulation then updates itself, according to the decisions, and writes information about its current state on each team's disk. The administrator then gives each team its disk, and the round repeats.
Network Configurations: Floppy disks are cumbersome, have limited storage capacity, and require all participants to be in close proximity. MTSs overcome these problems by running over a network.
With reference now to
The main station 2020 communicates with the central computer 2010 via a communication link 2030, for example, a modem or dedicated communication line. A plurality of stations 2040 are also connected to the central computer through respective communication lines 2050. Each station 2040 preferably comprises a personal or laptop computer such as one owned by the participant in the simulation. Preferably, each firm in the simulation (e.g., management training simulation) is controlled at one station 2040; however, a single station or user can control plural firms, or multiple stations can share responsibility for governing the activities of a single firm within the spirit of the present invention. The hardware arrangement 2000 of
An important case of this configuration occurs when the clients are not always connected to the server. This happens when running a simulation via the Internet and e-mail is used to transfer files. (Essentially, email replaces the floppy disks described in a previous configuration). To see the problem, consider the nk-landscape function. Recall that it keeps data in tables, and these tables can become very large. For this reason, the tables are not created in advance, but are generated as they are needed. This creates a problem if clients that are not permanently connected to the server need access to the SVF “on demand.” On demand access is preferable when modeling R&D in the manner previously described. The solution is to represent the nk-landscape function as an equation (see section B2.7.2 of HEC) and give the equation to each computer.
As this problem illustrates, an MTS can place a portion of the simulated business situation on the server and other portions with the clients. In fact, some portions can be duplicated on both. For example, a competitive industry simulation will often keep simulations of markets on the server. It will often keep the models of the firm(s) on the client computer(s). In this manner, the simulated business situation is spread over multiple computers in a network.
Those versed in the art will recognize that the preceding descriptions enable a wide variety of simulations. An MTS can include design objects representing objects, technologies, or other objects. The objects' can have any number of attributes, and the attributes can vary qualitatively, quantitatively, or in other ways. The attributes can be segmented, include system-set attributes, and have multiple properties. The simulated business situation can include simulations of R&D, manufacturing, or a competitive marketplace. It could also include simulated competitors, financial markets, accounting, supply curves for object characteristics, supply curves for manufacturing capacity, and simulations of other phenomena associated with business. Below are several flowcharts depicting some particularly useful configurations.
Flowchart for Evaluating a User's Design Performance
In accordance with one aspect of the present invention, the MTS can automatically evaluate the performance of a particular student's design or designs relative to predetermined criteria such as other student designs or bench mark levels established statistically or otherwise. With reference now to
At step 2100, the system obtains an attribute-characteristic representation of one or more designs from each of the students participating in the simulation. Such designs are obtained by completing a form that preferably is presented electronically on the display screen at the stations 2040. For example, a user interacts with the various fields displayed through the interface illustrated in
The central computer 2010 evaluates the designs posted by the students using a multipeaked value function, as at step 2110. The central computer outputs at step 2120 marketplace performance data to each student with respect to their respective designs. The marketplace performance data is communicated from the central computer over the communication lines 2050 to the stations 2040 and, more particularly, to the station which posted that particular design in the first place. Meanwhile, the designs of all students can be provided across communication line 2030 to the main station 2020 so that the instructor can review and monitor progress in the designs as the simulation proceeds. The central computer records in a memory each of the designs it obtains from the various stations 2040 along with the value computed by the multipeaked value function for the present round, as at step 2130.
Next, a determination is made as to whether the simulation is to continue, at step 2140. This simulation can continue for a predetermined number of rounds or until other predetermined criteria are satisfied. For example, a simulation may continue (1) until a set number of firms has gone bankrupt, (2) a certain number of rounds after a radical innovation was introduced into the marketplace, or (3) based on other criteria. In the event that the simulation is to continue, then, optionally, the multipeaked value function and/or the number of attributes and/or the domains of one or more attributes can be altered to simulate exogenous shifts in the marketplace. For example, when the multipeaked value function is altered, then the values for all objects in the simulation are affected. As another example, when the number of attributes and/or the domains of one or more attributes are altered, the simulation models the discovery of the new product, a shortage of raw materials, or government regulation.
At step 2160, the system obtains at step 2160 revised designs from the students along with other data respecting the simulation such as requests for reports, surveys, advertising budgets, budget allocations, revised production schedules, royalty payments, and the like. This information is obtained by posting a form as described above in connection with step 2100. The process flow then loops back to step 2110 and repeats so that the revised designs can be evaluated using the (optionally altered) multipeaked value function, with the results being output to the students and recorded at the central station.
In the event that the simulation is not to continue further, as tested at step 2140, the process flow instead branches to step 2180 at which step each student's design is automatically evaluated relative to predetermined criteria, as stated above.
This example was presented in the context of a simulated business situation, but this is not necessary. The student's task can simply be to design the best design object possible within a specified number of rounds or searches. The data collected by the simulation can include qualities of a students search. These qualities can be useful in studying cognitive psychology. Particularly, it is useful in studying the psychological processes described in the Section 4, “Embodiments that Make Two Great Improvement in MTSs.” These processes are categorization, covariation detection, and judgment.
Flowchart for Developing a User's Decision-making Skills
Turning now to
At step 2202, the simulation and firms are initialized; that is, the starting settings for firm , firm , . . . , firm [n] are established. The initial set up of a firm can be as shown in
In accordance with a description of modeling marketing (see Section 8, “Modeling Marketing MTSs”), the central computer receives from each station 2040 a filter setting which is used to guide the retrieval of marketplace data from the memory of the central computer. A filter setting can be provided by the participant by completing a form that includes the same type of information that is available through the marketplace data display shown in
Thereafter, a determination is made at step 2212 as to whether the simulation is to continue, substantially as described above in connection with step 2140 of
Regardless of whether the domain is updated, the process continues at step 2220 by obtaining from the user revised designs, production data, and other input from each firm, that is, from use station 2040. At step 2222, a determination is made as to whether a change in the value of any of the valid designs is appropriate. Such a change can reflect, for example, an exogenous shift in the product space such as a change in consumer preferences or inflation. In accordance with a feature of the invention, the central computer 2010 includes system-set attributes in the attribute-characteristic representation of the set of valid object designs. The system-set attributes can exhibit strong frustration with other attributes that are alterable by the user. When the system-set attributes are changed, the overall value of the product is impacted. This is manifested in the simulation as sudden change in the value of the object versus the value prior to the change. If the value is to be changed, then the system-set attribute is altered, and new values are assigned at step 2224.
Also, regardless of whether there is change in the value, a determination is next made at step 2226 as to whether the multipeaked value function itself is to be changed for a subsequent round. In some simulations, it may be desirable to utilize a different multipeaked function than an initial one, for example, a new function derived from the first multipeaked function. In such a simulation, a new function is assigned at step 2228. Regardless of whether such a new function is assigned, if the simulation is to continue, as was determined at step 2212, then a new round of marketplace data is generated using the present or current multipeaked value function, as described above in connection with step 2204. The process flow then proceeds as previously described, round-after-round, until an end of simulation condition is satisfied.
A management training system that can be used to implement the method of the present invention preferably includes a first computer having a processor and a memory and a network connection to a plurality of stations such as the arrangement shown in
Although the present invention has been fully described by way of example with reference to the accompanying drawings, it is to be noted here that various changes and modifications will be apparent to those skilled in the art. Therefore, unless such changes and modifications depart from the scope of the present invention, they should be construed as being included therein.
Key terms are listed in alphabetical order.
Innovation: In prior art MTSs, innovation either (1) increases the value of an aggregate trait, (2) expands the domain of attributes, (3) increases the number of attributes, or (4) adds an entirely new market to the learning session. Regardless of which effects are included, prior art MTSs use a variation of the following method to simulate innovation: For an example see: Thomas Pray and David Methe, “Modeling Radical Changes in Technology within Strategy-Oriented Business Simulations,” Simulation and Gaming, vol. 22 (March 1991): pp. 19-35. Firms attempt innovation by allocating capital to “research and development”. The allocation purchases a draw from a probability distribution. If the draw exceeds a predetermined threshold, the firm innovates. With this method, students have limited influence over the probability distribution and threshold parameter. The most potent means of influence is investment in research and development. Greater investment increases the probability of innovation.
The prior art method is quite flexible. The frequency of innovation, variation of innovative ability among firms, and firm and industry leaning curves, can all be modeled by adjusting the distribution and the threshold parameter. These capabilities make the method adequate for adding innovation to MTSs in which innovation is not an important aspect of the MTS (for example, MTSs that teach basic accounting, finance, and marketing in established markets). While useful in such MTSs, the prior art method of modeling innovation is inadequate for MTSs in which innovation is an important aspect (for example, MTSs that teach the management of innovation, technological advance, or knowledge). For such MTSs, the prior art is deficient for several reasons. These are:
For examples of prior art product value functions, see: Richard Teach, “Demand Equations for Business Simulations with Market Segments,” Simulation and Gaming, vol. 21 (December 1990): pp. 423-442 and Steven Gold and Thomas Pray, “Technological Change and Intertemporal Movements in Consumer Preferences in the Design of Computerized Business Simulations with Market Segmentation,” Developments in Business Simulations and Experiential Exercises, vol. 25 (1998): pp. 156-167.
Utilizing this method, product value is a monotonically decreasing function of the distance between two products. Products that have small distances (similar to the ideal) have higher values. Products that have large distances (dissimilar to the ideal) have lower values. This result arises because the distance functions do not represent frustration. The formula for calculating the distance between two points provides an example. Suppose a product class has four quantitative attributes. In prior art MTSs, the distance between a product and the ideal might be calculated as DISTANCE=(d1 2+2 2+ . . . +dn 2)0.5, where di=(Ii−ai) and Ii and ai represent the characteristics expressed by the ith attribute for the ideal product and a product produced by the student. By taking derivatives one can show that when the value function decreases monotonically with distance (1) an improvement in design in one dimension can never cause a decrease in the value of a product (i.e., no strong frustration) and (2) that the value function has only one optimum, which occurs when d1=d2= . . . =dn=0.
The mountain range metaphor helps to illuminate the qualities produced by this method. Because the distance value functions lack frustration, the product space topology of the prior art MTSs are Mount Fuji-like. The topology has a single peak that stands directly over the ideal product. As products become less like the ideal, their design quality decreases, producing broad, gradual slopes. This topology, has significant consequences, including:
These qualities do not exist in the present invention described in the foregoing specification. The strong frustration exhibited by multipeaked value functions prevents students from addressing each attribute independently. The strong frustration also prevents students from finding the optimal product via a sequence of products that (1) differ by incremental design changes and (2) produce a monotonically increasing sequence of product values. In the present inventive system and method, continuing improvement in product design eventually must require a more radical change in product design. Metaphorically, the product space topology of the present invention is multipeaked, as opposed to single peaked. Improving a product design that resides beneath the peak of a smaller mountain requires a simultaneous change of several product characteristics; one must travel to a different, taller mountain.
There is another significant quality of the prior art (and one that distinguishes the present system and method from the prior art). This distance value function creates a highly correlated product space. Information about the value of products provides considerable information about other products. The Mount Fuji topology illustrates this property. Once a student discovers the direction of design changes that “climb” the mountain, the student knows a great deal about the entire topology of the product space. This differs from the new method where information is correlated over small areas of the mountain range (among sets of incremental innovations) but uncorrelated across larger areas (areas containing radical innovations).
Some makers of prior art MTSs recognize that the prior art method produces highly reliable information. As a remedy, some prior art MTSs add a random error term to the marketplace information (for example, to the information found in a marketing report). The difference between this modeling of uncertainty and uncertainty in embodiments of the present invention is significant. The prior art adds informational uncertainty exogenously. In the new method, information reliability arises endogenously and produces the proper relationship between information reliability and innovation. Moreover, the problems of the prior art method are compounded by another practice of the prior art: making the reliability of information vary with investment in marketing research. For an example of this, see: Thomas Pray and David Methe, “Modeling Radical Changes in Technology within Strategy-Oriented Business Simulations,” Simulation and Gaming, vol. 22 (March 1991): pp. 19-35. More expensive marketing reports have more reliable information. In these MTSs, the reliability of information is a function of investment. In the new method, the information's reliability is, appropriately, a function of innovation and perspective.
Because distance value functions exhibit the qualities just described: (1) addressing the attributes independently; (2) optimizing of product value via a sequence of small changes in design that each improve value; and (3) highly correlated—they are suitable only for teaching the management of low uncertainty situations. These situations include, for example, pricing, designing, positioning, and promoting products in established markets (i.e., basic marketing). Prior art MTS largely are not suitable for teaching the management of high uncertainty situations. These situations include, for example, entrepreneurship, developing new core competencies, developing radical innovations, managing technological change, and reinventing one's business.
Finally, it should be noted that the prior art does not provide a method for measuring information and knowledge; relating these measures to the product space topology; or relating these measures to students' decisions. Furthermore, the prior art cannot relate the tasks demanded of students to cognitive functions. Because of these limitations, the prior art cannot usefully illuminate and analyze the role of knowledge and information in students' decisions, marketplace competition, and the simulated industry's dynamics. Moreover, for the same reasons, the prior art must teach through an indirect method, where students test a variety of ideas and, hopefully, induce an improved understanding of management. The prior art cannot teach through the direct method previously described.
|Cited Patent||Filing date||Publication date||Applicant||Title|
|US5041972||Apr 15, 1988||Aug 20, 1991||Frost W Alan||Method of measuring and evaluating consumer response for the development of consumer products|
|US5056792||Feb 6, 1990||Oct 15, 1991||Helweg Larsen Brian||Business education model|
|US5240419||Jul 15, 1991||Aug 31, 1993||Hughes Training, Inc.||Condition monitor request processing system|
|US5310349||Apr 30, 1992||May 10, 1994||Jostens Learning Corporation||Instructional management system|
|US5326270||Aug 29, 1991||Jul 5, 1994||Introspect Technologies, Inc.||System and method for assessing an individual's task-processing style|
|US5574889||Jan 27, 1994||Nov 12, 1996||Nissan Motor Co., Ltd.||Apparatus for selecting and evaluating design based on stored information|
|US5737581||Aug 30, 1995||Apr 7, 1998||Keane; John A.||Quality system implementation simulator|
|US5829983||May 2, 1995||Nov 3, 1998||Fujitsu Limited||System for carrying out educational management|
|US5887154||Jan 4, 1996||Mar 23, 1999||Hitachi, Ltd.||Business process simulation system|
|US5957699||Dec 22, 1997||Sep 28, 1999||Scientific Learning Corporation||Remote computer-assisted professionally supervised teaching system|
|US5974246||Sep 11, 1997||Oct 26, 1999||Nakazawa; Hiromu||Method of determining optimum product design parameters and system therefor|
|US6015348||Oct 18, 1996||Jan 18, 2000||Starwave Corporation||Scalable game server architecture|
|US6134539||Dec 22, 1998||Oct 17, 2000||Ac Properties B.V.||System, method and article of manufacture for a goal based education and reporting system|
|US6157808||Jul 16, 1997||Dec 5, 2000||Gpu, Inc.||Computerized employee certification and training system|
|US6236955||Jul 30, 1999||May 22, 2001||Gary J. Summers||Management training simulation method and system|
|US6292830 *||Aug 7, 1998||Sep 18, 2001||Iterations Llc||System for optimizing interaction among agents acting on multiple levels|
|US6408263||Jul 30, 1999||Jun 18, 2002||Gary J. Summers||Management training simulation method and system|
|US6430526 *||Dec 22, 1998||Aug 6, 2002||Intel Corporation||Computer processable interconnect topology|
|US6634950||Jul 27, 2001||Oct 21, 2003||Konami Corporation||Game system, game providing method, and information recording medium|
|US6755659||Jul 12, 2001||Jun 29, 2004||Access Technologies Group, Inc.||Interactive training system and method|
|US6931365||Dec 17, 2001||Aug 16, 2005||Purdue Research Foundation||Industry simulation environment|
|US6938048 *||Nov 14, 2001||Aug 30, 2005||Qgenisys, Inc.||Universal task management system, method and product for automatically managing remote workers, including automatically training the workers|
|US6939137||Dec 12, 2001||Sep 6, 2005||Cantor Fitzgerald, Lp||Method and system for training traders|
|US6941255||Jun 29, 2001||Sep 6, 2005||Object Reservoir, Inc.||Feature modeling in a finite element model|
|US6944596||Feb 23, 2000||Sep 13, 2005||Accenture Llp||Employee analysis based on results of an education business simulation|
|US6983236 *||Oct 11, 2000||Jan 3, 2006||Aprisa, Inc.||Method for system-constraint-based selection for design components|
|US20010039002||Feb 20, 2001||Nov 8, 2001||John Delehanty||System and method for implementing and managing training programs over a network of computers|
|US20030059760||Oct 8, 2002||Mar 27, 2003||Joe Tojek||Learning system and method for holding incentive-based learning|
|US20030084015||Aug 22, 2001||May 1, 2003||Beams Brian R.||Interactive simulations utilizing a remote knowledge base|
|US20040138934||Apr 18, 2003||Jul 15, 2004||General Electric Company||Controlling a business using a business information and decisioning control system|
|1||"Adaptation on Rugged Landscapes Generated by Iterated Local Interactions of Neighboring Genes," Marc Lipsitch, Santa Fe Institute, Santa Fe, New Mexico (in Proc. of the 4th Intl Conf. on Genetic Algorithms, San Mateo, CA, Morgan Kauffmann, 1991), pp. 128-135.|
|2||"Modeling Industry Life Cycles for Behavioral Research in Management and Economics," Gary J. Summers, Ph.D, New Orleans, Louisiana, (Social Science Research Electronic Publishing, Inc. May 28, 1998 www.ssm.com) pp. 1-44 plus abstract.|
|3||Advisory Action mailed Aug. 8, 2007 from co-pending U.S. Appl. No. 10/141,997.|
|4||Alden Wright, "Genetic Algorithms for Real Parameter Optimization," in Foundations of Genetic Algorithms, Gregory Rawlins, ed., pp. 205-218 (San Mateo, California: Morgan Kauffman Publishers, 1991).|
|5||Bernard Manderick, "Correlation Analysis," In Handbook of Evolutionary Computation, .sctn.B2.7.3, pp. B2.7:10-14 plus cover (New York: Oxford University Press, 1997).|
|6||Bernard Manderick, et al., "The Genetic Algorithm and the Structure of the Fitness Landscape," in Proc. of the 4th Intl Conf. on Genetic Algorithms, Belew and Brookes, eds., pp. 143-150 (San Mateo, California: Morgan Kauffmann Publishers, 1991).|
|7||Business Simulation [online]. Strategic Management Group, Inc. 1999 [ret. Sep. 30, 1999].|
|8||Campbell, T.L. Optimal Decision Making in a Complex Business Simulation, IEEE proc. Of 22.sup.nd Annual Hawaii int'l. Conf. On System Sciences, Jan. 1989. vol. 3. pp. 822-831.|
|9||Capitalism [online]. Games Domain Review, 1996 [retrieved on Sep. 30, 1999].|
|10||Capitalism Manual User's Guide (1995), Trevor Chan, Interactive Magic, Morrisville, North Carolina, Chapter 5; pp. 5-1 through 5-12.|
|11||Capitalism Product Description [online]. Interactive Magic, 1996 [ret. On Sep. 30, 1999].|
|12||Daniel Kahneman et al., "California of Probabilities: The State of the Art," in Judgement Under Uncertainty; Heuristics and Biases, Chapter 22, pp. 306-334 plus cover (1982).|
|13||Darrell Whitley, "Permutations," in Handbook of Evolutionary Computation, Thomas Back et al., ed., .sctn.C1.4, pp. C1.4, pp. C1.4:1-8 plus cover (1997).|
|14||Dennis Jennings, et al., "Informal Covariation Assessment: Data-Based versus Theory-Based Judgments," in Judgment Under Uncertainty: Heuristics and Biases, Chapter 15, pp. 211-230 (New York: Cambridge University Press, 1982).|
|15||Donald Frey, "The New Dynamism (Part 1)," in Interfaces, vol. 24, No. 2, pp. 87-91 (Mar.-Apr. 1994).|
|16||Gary J. Summers, "Analyzing Managers' Judgments and Decisions with an Educational Business Simulation," in Developments in Business Simulation and Experiential Learning, Sandra Morgan and Diana Page, eds., vol. 26, pp. 58-64 plus cover (1999).|
|17||Gary J. Summers, "Industrial Dynamics: An Evolutionary Model for an Interactive Simulation," from Dissertation Submitted to the Graduate School of Northwestern University, pp. 1-166 (Jun. 1997).|
|18||Gary J. Summers, "Modeling Innovation as a Process of Design in Educational Business Simulations," in Developments in Business Simulation and Experiential Learning, Sandra Morgan and Diana Page, eds., vol. 26, pp. 146-152 plus cover (1999).|
|19||James Brian Quinn, "Managing Innovation: Controlled Chaos," In Harvard Business Review, pp. 73-84 (May-Jun. 1985).|
|20||James M. Utterback, "Innovation and Industrial Evolution," In Mastering the Dynamics of Innovation-How Companies Can Seize Opportunities in Face of Technological Change, Chapter 4, pp. 79-102 plus cover (1994).|
|21||Jose Lobo et al., "Landscapes: A Natural Extension of Search Theory," from Santa Fe Inst. Website, pp. 1-33 plus abstract and cover (1999).|
|22||Kate Juliff, "The Packing Problem," in Handbook of Evolutionary Computation, Thomas Back et al., ed., .sctn.F1.7, pp. F1.7:1-5 plus cover (1997).|
|23||Lowell W. Steele, "Nurturing Innovation," in Managing Technology: The Strategic View, Chapter 11, pp. 263-288 (New York: McGraw-Hill, 1989).|
|24||Office Action mailed May 8, 2007 (made final) from co-pending U.S. Appl. No. 10/141,997.|
|25||Peter Drucker, "Entrepreneurial Management," in Innovation and Entrepreneurship-Practice and Principles, Chapter 12, pp. 143-176 (New York Harper & Row, 1985).|
|26||Peter Drucker, "The Innovative Organization," In Management: Tasks, Responsibilities, Practices, Chapter 61, pp. 782-803 (New York: Harper Collins, 1973).|
|27||Philip Auerswald et al., "The Production Recipes Approach to Modeling Technological Innovation: An Application to Learning by Doing," from Santa Fe Inst. Website, pp. 1-42 plus Tables 1-10, Figs. 1-24, cover and abstract (1998).|
|28||Powerim Constructor 2.5 [online]. Vison Works 1999 [ret. Sep. 30, 1999].|
|29||Powersim [online] Powersim 1998 [retrieved on Sep. 30, 1999].|
|30||Precha Thavikulwat, "Modeling the Human Component in Computer-Based Business Simulations," in Simulation and Gaming, vol. 22, No. 3, pp. 350-359 (Sep. 1991).|
|31||Precha Thavikulwat, "Product Quality in Computerized Business Simulations," in Simulation and Gaming, vol. 23, No. 4, pp. 431-441 (Dec. 1992).|
|32||Ralf Bruns, "Scheduling," in Handbook of Evolutionary Computation, Thomas Back et al., ed., .sctn.F1.5, pp. F1.7:1-9 plus cover (1997).|
|33||Request for Reconsideration filed Jul. 9, 2007 from co-pending U.S. Appl. No. 10/141,997.|
|34||Richard D. Teach, "Demand Equations for Business Simulations With Market Segments," in Simulation and Gaming, vol. 21, No. 4, pp. 423-442 (Dec. 1990).|
|35||Richard R. Nelson et al., "Further Analysis of Search and Selection," In An Evolutionary Theory of Economic Change, Chapter 11 (Introduction and Part 1), pp. 246-262 plus cover (1982).|
|36||Richard R. Nelson, "The Role of Knowledge In R&D Effciency," In The Quarterty J. of Econ., pp. 453-470 (1982).|
|37||Robyn M. Dawes, Rational Choice In An Uncertain World, .sctn..sctn.2. 1-2.3 (pp. 22-31), .sctn.6.5 (pp. 121-125), and .sctn..sctn.7.1-7.8 (pp. 128-143), plus cover and abstract (1988).|
|38||Set of claims subject to rejection in U.S. Appl. No. 10/141,997 , 2007.|
|39||Steven C. Gold et al., "Modeling Demand in Computerized Business Simulations," in Guide to Business Gaming and Experiential Learning, Association for Business Simulation and Experiential Learning, Jim W. Gentry, ed., Chapter 8, pp. 117-138 plus cover (East Brunswick: Nichols/GP Publishing, 1990).|
|40||Steven C. Gold et al., "Technological Change and Intertemporal Movements in Consumer Preferences in the Design of Computerized Business Simulations with Market Segmentation," in Developments in Business Simulation and Experiential Learning, Nancy H. Leonard and Sandra W. Morgan, eds., vol. 25, pp. 156-167 (1998).|
|41||Steven C. Gold, "Modeling Short-Run Cost and Production Functions in Computerized Business Simulations," in Simulation and Gaming, vol. 23, No. 4, pp. 417-430 (Dec. 1992).|
|42||Steven Gold and Thomas Pray, "The Production Frontier: Modeling Production in Computerized Business Simulations," in Simulation and Games, vol. 20, No. 3, pp. 300-318 (Sep. 1989).|
|43||Steven Gold et al., "Modeling Market-and Firm-Level Demand Functions in Computerized Business Simulations," In Simulations and Games, vol. 15, No. 3, pp. 346-363 (Sep. 1984).|
|44||Stuart A. Kauffman, "The Evolution of Economic Webs," in The Economy As An Evolving Complex System, Philip Anderson, Proc. of the Evolutionary Paths of the Global Econ. Workshop Held in Santa Fe, New Mexico, Kenneth Arrow and David Pines, eds., pp. 125-146 (Sep. 1987).|
|45||Stuart Kauffman et al., "Optimal Search on a Technology Landscape," from Santa Fe Inst. Website, pp. 2-40 plus abstract and table of contents (1998).|
|46||Stuart Kauffman et al., "Technological Evolution and Adaptive Organizations," in Complexity, vol. 1, No. 2, pp. 26-43 plus table of contents (1995).|
|47||Stuart Kauffman, "The Structure of Rugged Fitness Landscapes," In The Origins of Order: Self-Organization and Selection in Evolution, Chapter 2, pp. 33-67 (New: York Oxford University Press, 1993).|
|48||Thomas Back, Evolutionary Algorithms in Theory and Practice, .sctn.3.4, pp. 143-144 and cover (New York: Oxford University Press, 1996).|
|49||Thomas F. Pray et al., "Modeling Radical Changes in Technology Within Strategy-Oriented Business Simulations," in Simulation and Gaming, vol. 22, No. 1, pp. 19-35 (Mar. 1991).|
|50||Volker Nissen, Management Applications and Other Classical Optimization Problems, in Handbook of Evolutionary Computation, Thomas Back et al., .sctn.F1.2, pp. F1.2:1-50 plus cover (1997).|
|Citing Patent||Filing date||Publication date||Applicant||Title|
|US7580853 *||Apr 13, 2007||Aug 25, 2009||Electronic Entertainment Design And Research||Methods of providing a marketing guidance report for a proposed electronic game|
|US7729942 *||Sep 29, 2006||Jun 1, 2010||At&T Intellectual Property I, L.P.||Consumer targeting methods, systems, and computer program products using multifactorial marketing models|
|US7835893||Sep 3, 2003||Nov 16, 2010||Landmark Graphics Corporation||Method and system for scenario and case decision management|
|US7933673 *||Jul 15, 2009||Apr 26, 2011||I2 Technologies Us, Inc.||Demand breakout for a supply chain|
|US8024218 *||Mar 23, 2006||Sep 20, 2011||Sap Ag||Method and apparatus for determining the product marketability utilizing a percent coverage|
|US8060393||Jul 1, 2009||Nov 15, 2011||Electronic Entertainment Design And Research||Characteristics of players systems and methods for analyzing electronically embodied games|
|US8195533 *||Dec 8, 2008||Jun 5, 2012||Amazon Technologies, Inc.||Self healing andon system and method|
|US8209202 *||May 1, 2006||Jun 26, 2012||Landmark Graphics Corporation||Analysis of multiple assets in view of uncertainties|
|US8315897||Nov 7, 2011||Nov 20, 2012||Electronic Entertainment Design & Research||Systems and methods for evaluating, classifying and predicting game trends using clustered pattern recognition|
|US8392455 *||Jul 2, 2009||Mar 5, 2013||Palo Alto Research Center Incorporated||Multi-interval heuristics for accelerating target-value search|
|US8458000||May 17, 2012||Jun 4, 2013||Landmark Graphics Corporation||Analysis of multiple assets in view of functionally-related uncertainties|
|US8521575||Aug 24, 2012||Aug 27, 2013||Electronic Entertainment Design And Research||Systems and methods for classifying computer video game genres utilizing multi-dimensional cloud chart|
|US8589125 *||Oct 3, 2008||Nov 19, 2013||Ihi Corporation||Product design support system and method for simulating a prototype of a design object|
|US8620776||Jun 4, 2012||Dec 31, 2013||Amazon Technologies||Self healing andon system and method|
|US8712747||Nov 15, 2010||Apr 29, 2014||Landmark Graphics Corporation||Decision management system and method|
|US8886550||Aug 16, 2013||Nov 11, 2014||Electronic Entertainment Design And Research||Systems and methods for classifying computer video game genres utilizing multi-dimensional cloud chart|
|US9508051 *||Mar 15, 2013||Nov 29, 2016||Bmc Software, Inc.||Business development configuration|
|US9805148 *||Aug 18, 2014||Oct 31, 2017||The United States Of America, As Represented By The Secretary Of The Navy||Modeling and simulation analytical reply method|
|US20040220790 *||Sep 3, 2003||Nov 4, 2004||Cullick Alvin Stanley||Method and system for scenario and case decision management|
|US20050197821 *||Aug 20, 2004||Sep 8, 2005||Korea Advanced Institute Of Science And Technology||System and a method for workflow system (engine) based workflow model simulation|
|US20070156421 *||Dec 29, 2005||Jul 5, 2007||Brian Klapper||Training using simulated work environment|
|US20070226040 *||Mar 23, 2006||Sep 27, 2007||Ashutosh Kumar||Product market determination|
|US20070244742 *||Apr 13, 2007||Oct 18, 2007||Electronic Entertainment Design And Research||Systems and Methods For Analyzing Electronically Embodied Games|
|US20080082411 *||Sep 29, 2006||Apr 3, 2008||Kristina Jensen||Consumer targeting methods, systems, and computer program products using multifactorial marketing models|
|US20080120148 *||May 1, 2006||May 22, 2008||Keshav Narayanan||Analysis of multiple assets in view of uncertainties|
|US20090055230 *||Aug 22, 2007||Feb 26, 2009||Craig Thomas Ellrod||Process for validating and proving technical marketing and competitive analysis from hands-on lab testing and publicly available research information|
|US20090222310 *||Feb 29, 2008||Sep 3, 2009||Microsoft Corporation||Techniques to allocate project resources|
|US20090276275 *||Jul 15, 2009||Nov 5, 2009||Brown Richard W||Demand breakout for a supply chain|
|US20100003645 *||Jul 2, 2008||Jan 7, 2010||Moresteam.Com Llc||Education method and tool|
|US20100021872 *||Jul 22, 2008||Jan 28, 2010||Moresteam.Com Llc||Process of learning process improvement techniques|
|US20100023381 *||Jul 1, 2009||Jan 28, 2010||Short Gregory T||Characteristics Of Players Systems & Methods For Analyzing Electronically Embodied Games|
|US20100250200 *||Oct 3, 2008||Sep 30, 2010||Ihi Corporation||Product design support system and method|
|US20110004625 *||Jul 2, 2009||Jan 6, 2011||Palo Alto Research Center Incorporated||Multi-Interval Heuristics For Accelerating Target-Value Search|
|US20110060573 *||Nov 15, 2010||Mar 10, 2011||Alvin Stanley Cullick||Decision Management System and Method|
|US20110066476 *||Sep 15, 2009||Mar 17, 2011||Joseph Fernard Lewis||Business management assessment and consulting assistance system and associated method|
|US20110112880 *||Sep 21, 2010||May 12, 2011||Ebay Inc.||Allocation of common resources in an entity|
|US20120221368 *||Feb 28, 2011||Aug 30, 2012||Sui Yang||Procedure planning tool for office appointments management|
|US20120221380 *||Feb 28, 2011||Aug 30, 2012||Bank Of America Corporation||Teller Readiness Simulation|
|US20140046709 *||Aug 8, 2012||Feb 13, 2014||Mark Clement Kwapiszeski||Methods and systems for evaluating technology assets|
|US20140122144 *||Nov 1, 2012||May 1, 2014||Vytas Cirpus||Initiative and Project Management|
|US20140278818 *||Mar 15, 2013||Sep 18, 2014||Bmc Software, Inc.||Business development configuration|
|US20160048613 *||Aug 18, 2014||Feb 18, 2016||United States Of America As Represented By The Secretary Of The Navy||Modeling and Simulation Analytical Reply Method|
|U.S. Classification||703/22, 434/107, 703/6, 706/925, 706/920|
|International Classification||G09B19/18, G09B5/00, G06Q30/00, G06Q10/00, G06Q99/00, G06F9/45|
|Cooperative Classification||G06Q30/02, Y10S706/925, G06Q99/00, G09B5/00, Y10S706/92, G09B19/18, G06Q10/10|
|European Classification||G06Q30/02, G06Q10/10, G09B19/18, G06Q99/00, G09B5/00|
|Mar 27, 2011||FPAY||Fee payment|
Year of fee payment: 4
|Aug 25, 2015||FPAY||Fee payment|
Year of fee payment: 8
|Jan 19, 2016||AS||Assignment|
Owner name: GULA CONSULTING LIMITED LIABILITY COMPANY, DELAWAR
Free format text: MERGER;ASSIGNOR:POLIMATRO SOFTWARE LIMITED LIABILITY COMPANY;REEL/FRAME:037526/0386
Effective date: 20150826